Welcome to my Luma Profile!

Blog Posts

I send an email recap of my week on Sunday evenings. 

The Weekly Update

Jamie XX — Rejkjavik

Purity Ring — Montreal

Skillex — Shanghai

ZHU — Japan

Alison Wonderland — Los Angeles

RÜFÜS DU SOL - Joshua Tree

Favorite DJ Sets

Events

My Website

Luma Homepage

Around the Web

Cheap

Spring Special

Big Bundle

Running Seminar — How to Run Without Hurting Yourself

Breathwork Workshop Test

Example Member Only Event

Clubhouse Event Pages with lu.ma/clubhouse

Series Example

Test Ticket!

Test Visibility

Growth Session

Happy Hour

Mindful Masters: Energy Healing

Luma Beta Overview

Build Club 1 — Making Games in the Browser

No Show Event

Event Chat Demo

Test Event

Luma Product Overview

tiptap Conversation

Test Water

Testy Test

Public Event

West Village Dinner Party

[The Mintable] - August - Manager Foundation - Cohort 4N - EMEA

Victor Pontis

At Luma / Glow, we run most of our infrastructure on AWS. Our main services are a database on Aurora and a website and API running on ECS. 
While AWS does a good job of running these jobs, we've had a really hard time running cron jobs, data backfill tasks, and one-off scripts on AWS. That's where Airplane comes in. 
Airplane.dev Overview
Airplane is awesome because it gives you nice infrastructure around running your own code. 
We give Airplane our main backend Docker image and then every Airplane task calls that one Dockerfile with a different command. This is nice since I didn't have to learn anything special to use Airplane, I can use what we are already deploying to AWS. 
This did take some time to set up the first time since I was a very early Airplane customer. But it sounds like the onboarding has gotten a lot easier. 
Below is what our Airplane dashboard looks like:

I've redacted the task names, but you can think about them in a few different categories:
Support Scripts
Before Airplane, if we wanted to delete a user or add a flag, we would have to SSH into one of our production machines. This was really annoying in that it was error prone, we didn't have auditing, and it required an engineer with access to production. 
Airplane lets you run scripts from their dashboard. And the logging if f'ing incredible. You can always click into an old script and see who ran it, what the parameters were, and look at the logs. You even get an email / Slack if the task fails!
I really really love the built in logging. Managing scheduled tasks and one off scripts on AWS is a nightmare and I was planning on building a whole system around this until I found Airplane!

Here is what one of the task detail pages looks like. And if I click into any run I can see exactly what happened. It's incredible. 
Tip for Airplane — It would be sweet if I could define a result of the the task and then see it in this list. 
Cron Jobs
For Glow, we use this to periodically backfill data that changes on a daily basis. For Luma, we use this to send guest email reminders for events. 
Airplane has a pretty nice scheduler that lets you define the task either in Cron format or a more humane way. 

Backfilling Data 
For Glow, we are pulling in and processing a ton of data. While new data gets added by a service on AWS, Airplane gives us an incredible interface for running tasks to populate old data. 
Again this benefits from the increased logging I mentioned above. If we are backfilling in a few months of data, I can trigger a month at a time since Airplane makes it easy for me to see the status of previous jobs. 

Thank you to the Airplane team!

<div class="mirror-content">​At <a href="https://lu.ma/" target="_blank" rel="nofollow noopener">Luma</a> / <a href="https://glow.app/" target="_blank" rel="nofollow noopener">Glow</a>, we run <a href="https://lu.ma/p/zPamspcICt0f3q9/Lumas-Tech-Stack" target="_blank" rel="nofollow noopener">most of our infrastructure</a> on AWS. Our main services are a database on Aurora and a website and API running on ECS. ​While AWS does a good job of running these jobs, we&#x27;ve had a really hard time running cron jobs, data backfill tasks, and one-off scripts on AWS. That&#x27;s where Airplane comes in. <h1>​<a href="http://Airplane.dev" target="_blank" rel="nofollow noopener">Airplane.dev</a> Overview</h1>​Airplane is awesome because it gives you nice infrastructure around running your own code. ​We give Airplane our main backend Docker image and then every Airplane task calls that one Dockerfile with a different command. This is nice since I didn&#x27;t have to learn anything special to use Airplane, I can use what we are already deploying to AWS. ​This did take some time to set up the first time since I was a very early Airplane customer. But it sounds like the onboarding has gotten a lot easier. ​Below is what our Airplane dashboard looks like:<img alt="Image" class="image" src="https://cdn.lu.ma/cdn-cgi/image/format=auto,dpr=2,quality=85,width=800/editor-images/1f/f4a2bfc3-66c1-4a83-bf0c-309208bab949" style="display:block;max-width:100%;max-height:400px;margin:24px auto"/>​I&#x27;ve redacted the task names, but you can think about them in a few different categories:<h2>​Support Scripts</h2>​Before Airplane, if we wanted to delete a user or add a flag, we would have to SSH into one of our production machines. This was really annoying in that it was error prone, we didn&#x27;t have auditing, and it required an engineer with access to production. ​Airplane lets you run scripts from their dashboard. And the logging if f&#x27;ing incredible. You can always click into an old script and see who ran it, what the parameters were, and look at the logs. You even get an email / Slack if the task fails!​I really really love the built in logging. Managing scheduled tasks and one off scripts on AWS is a nightmare and I was planning on building a whole system around this until I found Airplane!<img alt="Image" class="image" src="https://cdn.lu.ma/cdn-cgi/image/format=auto,dpr=2,quality=85,width=460/editor-images/ot/4e5e11e2-21df-4d42-8b53-52825e44bd0c" style="display:block;max-width:100%;max-height:100%;margin:24px auto"/>​Here is what one of the task detail pages looks like. And if I click into any run I can see exactly what happened. It&#x27;s incredible. ​Tip for Airplane — It would be sweet if I could define a result of the the task and then see it in this list. <h2>​Cron Jobs</h2>​For Glow, we use this to periodically backfill data that changes on a daily basis. For Luma, we use this to send guest email reminders for events. ​Airplane has a pretty nice scheduler that lets you define the task either in Cron format or a more humane way. <img alt="Image" class="image" src="https://cdn.lu.ma/cdn-cgi/image/format=auto,dpr=2,quality=85,width=800/editor-images/ie/72982485-6336-4b20-aba2-a8964f8bc535" style="display:block;max-width:100%;max-height:400px;margin:24px auto"/><h2>​Backfilling Data </h2>​For Glow, we are pulling in and processing a ton of data. While new data gets added by a service on AWS, Airplane gives us an incredible interface for running tasks to populate old data. ​Again this benefits from the increased logging I mentioned above. If we are backfilling in a few months of data, I can trigger a month at a time since Airplane makes it easy for me to see the status of previous jobs. ​​Thank you to the Airplane team!</div>

How we use Airplane to simplify our tech infrastructure

Current Issues
The current NFT spec is pretty bad for a few reasons:
every NFT requires multiple accounts (3+)
the token account address can change (and the token account is not always the associated token account of the current NFT owner)
the NFT holder can transfer the NFT to a diff token account which can make any existing escrow / marketplace accounts invalid
it’s hard to get all of the NFTs for a given wallet
the editions part of the spec is very confusing
it’s hard to group collections — not built into the original version of the spec and slow to query
it’s really hard to get all of the activity (transfers / etc) for one NFT since one tx may include the mint while another tx includes the token account
This all stems from the first problem — an NFT is a bunch of different accounts but it should really be just one account.
A Path Forwards
NFToken + One Account
In this new spec, we will have an NFToken program that we will use instead of the Tokenkeg program.
An NFT will be represented by an account that is a PDA of the NFToken program and it will have the following fields:
address — the ID of the NFT will be it’s address on the Solana blockchain
data — the rest of the fields will be stored in the dataholder — address (notice how I didn’t call this owner)
delegate — has ability to transfer the NFT to someone else
metadata_url — max 128 characters, you will want to use short URLs. the URL will store offchain metadata‚ spec https://docs.opensea.io/docs/metadata-standards
name — max 32 utf-8 characters
image_url — max 128 characters, name and image_url are nice to have on chain since this is the basics of what’s required for most NFTs
collection — points to the address of the collection (explained below)
creator / update_authority — the account used to sign the message to mint this NFT
Common Queries
Get All NFTs for a Wallet
getProgramAccountsprogram = NftToken
memcmp: holder = wallet
Transfer NFT
NFToken will have a transfer instruction that can be signed by either the holder or the delegate. After the transfer ix, the delegate is zero’d out.
Set Delegate
The holder can sign an instruction to set the delegate of the NFT.
Mint New NFT as Creator
To create a new NFT, you will call the mint instruction on the NFToken program and you pass in all of the data for the newly created NFT.
Update NFT
The update_authority has permissions to update the following fields:
name
image_url
metadata_url
Marketplace Actions
List NFT
To list an NFT, you will:
update the NFT delegate to the marketplace custody account
create a listing account that represents the sale information
An NFT will still show up in your wallet while it is listed on the NFT marketplace.
Update Listing
To update a listing, you update the listing account on the NFT marketplace. You don’t need to interact with NFToken
Cancel Listing
You can delete the listing account which should zero out the delegate field.
You can also zero out the delegate field yourself without getting permission from the NFT marketplace. They would have to update the listing account on their side and delete it.
Execute Sale
The delegate which is controlled by the marketplace can transfer the holder field after which the delegate field is zerod out.
So the NFT marketplace will make the transfer after seeing that a sale is executable.
Collections
We want to support collections from the inception of this protocol. While we support collections, not every NFT needs to be on a collection.
The NFToken program will also create PDAs that represent collections:
address — this is the ID of the collection
dataname — 32 characters
image_url — 32 characters
metadata_url — 128 characters
update_authority
Update Collection
The update_authority can update the name and metadata_url for the collection
Add NFT to Collection
To add an NFT, we will need signatures from the NFT update_authority and the collection update_authority.
Remove NFT from Collection
The collection update_authority can remove an NFT from its collection. This does not require permission from the NFT update_authority.
Further Work
If you're interested in pushing this spec forwards, @-me on Twitter.

<div class="mirror-content"><h2>​Current Issues</h2>​The current NFT spec is pretty bad for a few reasons:<ul><li>​every NFT requires multiple accounts (3+)</li><li>​the token account address can change (and the token account is not always the associated token account of the current NFT owner)</li><li>​the NFT holder can transfer the NFT to a diff token account which can make any existing escrow / marketplace accounts invalid</li><li>​it’s hard to get all of the NFTs for a given wallet</li><li>​the editions part of the spec is very confusing</li><li>​it’s hard to group collections — not built into the original version of the spec and slow to query</li><li>​it’s really hard to get all of the activity (transfers / etc) for one NFT since one tx may include the mint while another tx includes the token account</li></ul>​This all stems from the first problem — an NFT is a bunch of different accounts but it should really be just one account.<h1>​A Path Forwards</h1><h2>​<code>NFToken</code> + One Account</h2>​In this new spec, we will have an <code>NFToken</code> program that we will use instead of the <code>Tokenkeg</code> program.​An NFT will be represented by an account that is a PDA of the <code>NFToken</code> program and it will have the following fields:<ul><li>​<code>address</code> — the ID of the NFT will be it’s address on the Solana blockchain</li><li>​<code>data</code> — the rest of the fields will be stored in the data<ul><li>​<code>holder</code> — address (notice how I didn’t call this <code>owner</code>)</li><li>​<code>delegate</code> — has ability to transfer the NFT to someone else</li><li>​<code>metadata_url</code> — max 128 characters, you will want to use short URLs. the URL will store offchain metadata‚ spec <a href="https://docs.opensea.io/docs/metadata-standards" target="_blank" rel="nofollow noopener">https://docs.opensea.io/docs/metadata-standards</a></li><li>​<code>name</code> — max 32 utf-8 characters</li><li>​<code>image_url</code> — max 128 characters, <code>name</code> and <code>image_url</code> are nice to have on chain since this is the basics of what’s required for most NFTs</li><li>​<code>collection</code> — points to the address of the collection (explained below)</li><li>​<code>creator</code> / <code>update_authority</code> — the account used to sign the message to mint this NFT</li></ul></li></ul><h2>​Common Queries</h2>​Get All NFTs for a Wallet<ul><li>​<code>getProgramAccounts</code><ul><li>​<code>program = NftToken</code></li><li>​<code>memcmp: holder = wallet</code></li></ul></li></ul>​Transfer NFT​<code>NFToken</code> will have a transfer instruction that can be signed by either the <code>holder</code> or the <code>delegate</code>. After the transfer ix, the <code>delegate</code> is zero’d out.​Set Delegate​The <code>holder</code> can sign an instruction to set the <code>delegate</code> of the NFT.​Mint New NFT as Creator​To create a new NFT, you will call the <code>mint</code> instruction on the <code>NFToken</code> program and you pass in all of the data for the newly created NFT.​Update NFT​The <code>update_authority</code> has permissions to update the following fields:<ul><li>​<code>name</code></li><li>​<code>image_url</code></li><li>​<code>metadata_url</code></li></ul><h2>​Marketplace Actions</h2>​List NFT​To list an NFT, you will:<ul><li>​update the NFT <code>delegate</code> to the marketplace custody account</li><li>​create a <code>listing</code> account that represents the sale information</li></ul>​An NFT will still show up in your wallet while it is listed on the NFT marketplace.​Update Listing​To update a listing, you update the <code>listing</code> account on the NFT marketplace. You don’t need to interact with <code>NFToken</code>​Cancel Listing​You can delete the <code>listing</code> account which should zero out the <code>delegate</code> field.​You can also zero out the <code>delegate</code> field yourself without getting permission from the NFT marketplace. They would have to update the <code>listing</code> account on their side and delete it.​Execute Sale​The <code>delegate</code> which is controlled by the marketplace can <code>transfer</code> the <code>holder</code> field after which the <code>delegate</code> field is zerod out.​So the NFT marketplace will make the transfer after seeing that a sale is executable.<h2>​Collections</h2>​We want to support collections from the inception of this protocol. While we support collections, not every NFT needs to be on a collection.​The <code>NFToken</code> program will also create PDAs that represent collections:<ul><li>​<code>address</code> — this is the ID of the collection</li><li>​<code>data</code><ul><li>​<code>name</code> — 32 characters</li><li>​<code>image_url</code> — 32 characters</li><li>​<code>metadata_url</code> — 128 characters</li><li>​<code>update_authority</code></li></ul></li></ul>​Update Collection​The <code>update_authority</code> can update the <code>name</code> and <code>metadata_url</code> for the collection​Add NFT to Collection​To add an NFT, we will need signatures from the NFT <code>update_authority</code> and the collection <code>update_authority</code>.​Remove NFT from Collection​The collection <code>update_authority</code> can remove an NFT from its collection. This does not require permission from the NFT <code>update_authority</code>.<h2>​Further Work</h2>​If you&#x27;re interested in pushing this spec forwards, @-me <a href="https://twitter.com/victorpontis" target="_blank" rel="nofollow noopener">on Twitter</a>.</div>

Proposal: New Solana NFT Spec

Solana often takes a while for people to wrap their heads around. I find this happens for a few reasons:
While there is some good content, it's scattered across many websites and blogs.
Solana is often explained in relation to Ethereum. But you don't need to understand Ethereum to understand Solana. This just confuses things.
There are some concepts in Solana (like the NFT token model) which are more complicated than they need to be.
As with everything in crypto, the majority of explainers are non-technical that use hard to understand metaphors.
Solana is a Key / Value Database
I've spent a few months banging my head at this and I think I have a mental model that makes sense now...
Solana is a huge key  value data database. Each Solana node stores this key  value database in memory (and is why nodes have to be so big).
In Solana nomenclature, the key is called an address and the value is called the account. Each address points to an account where data is stored.
And everything on Solana is an address / account pair. For example, each program has an address and stores the code for the program at the associated account. (It's actually a bit more complex than this but we don't need to get into that...)

Regular Accounts and Storage Accounts
Now let’s talk about addresses. On Solana an address is a 256 bit number that is most often encoded with base58, so it looks something like this CXk5NCtYva7h4BcGXg5BDBDtZDwgBuWZcwmWt5ReY1yA.
There are two different types of accounts — let’s call them Regular accounts and Storage accounts.

Regular Accounts
Regular accounts are what you think of when you think of crypto addresses.
Bitcoin, Ethereum, and Solana all use Elliptic Curve cryptography which allows a public  private keypair to sign messages. If you have a public  private keypair, you can sign a message with your private key and other people can verify the message is signed correctly.
This public / private keypair mechanism allows wallets to sign transactions that authorize some action. The Bitcoin code won’t execute a transaction that sends Bitcoin unless that account has signed the transaction.
On Solana, the regular account corresponds to the public key. You hold the private key securely in your wallet and you can share your public key.
In this system, a public key is a 256 bit number which lies on an elliptic curve. So a valid Solana Regular Account is one which lies on the elliptic curve.
Regular accounts are used for a couple things:
Wallets — If you want to have an account that stores SOL, you will create a regular account. SOL is stored directly in your wallet. (Tokens / NFTs are stored on Storage Accounts associated with your wallet.)
Programs — Every program is a regular account. It stores data in the account that is executed when the program is called.

Storage Accounts (PDAs)
Storage accounts allow you to store data associated with programs. A program can do anything from storing the state of an NFT listing, allowing an escrow between two parties, to acting as a domain registry.
You know an account is a storage account because its address is not on the elliptic curve that we mentioned above. That means a storage account address is not part of a public / private keypair, it’s just a number.
Regular accounts can sign messages with their private key to authorize transactions. Storage accounts can’t do this since they don’t have a private key. Instead storage accounts are linked to a controller program.
So if you are creating a program that will take a poll of where everyone wants to go on vacation, that program will store information about the different vacation destinations and votes at storage programs that are linked to the program.

How NFTs Work
If you ask me, the NFT spec in Solana is as complicated as possible (and I really want to rewrite it...). You can’t actually understand the NFT spec without understanding the coin spec.
A coin (aka a token) on Solana has a few characteristics:
An ID — This is called the Mint and is stored at the Mint account. The Mint account is a Storage Account that is controlled by the Solana Token Program.
Metadata — Name, symbol, price, etc. This information is stored in accounts on chain and places off chain.
Total Supply — Let’s say there are 1 million VIC coin. This VIC could be stored on one account or spread out across a bunch of different accounts. Each account stores a certain number of VIC and points to the owner. So I could have multiple accounts that store VIC. My overall balance is the sum of the VIC in these accounts.
Now that we understand how coins work, we turn to NFTs. An NFT is just a coin with a supply of 1. So each NFT has the following accounts (which are all Storage Accounts):
Mint — This is the ID of the NFT and stores some basic metadata.
Metadata — This has even more Metadata about the NFT. For historical reasons, it’s another storage account...
Token — This stores who owns the NFT. This can be super confusing since the address of the token account can change over time.
So if you own an NFT, what that means is the Token account for the NFT says that you are the owner.

<div class="mirror-content">​Solana often takes a while for people to wrap their heads around. I find this happens for a few reasons:<ul><li>​While there is some good content, it&#x27;s scattered across many websites and blogs.</li><li>​Solana is often explained in relation to Ethereum. But you don&#x27;t need to understand Ethereum to understand Solana. This just confuses things.</li><li>​There are some concepts in Solana (like the NFT token model) which are more complicated than they need to be.</li><li>​As with everything in crypto, the majority of explainers are non-technical that use hard to understand metaphors.</li></ul><h1>​Solana is a Key / Value Database</h1>​I&#x27;ve spent a few months banging my head at this and I think I have a mental model that makes sense now...​Solana is a huge key value data database. Each Solana node stores this key value database in memory (and is why nodes have to be so big).​In Solana nomenclature, the key is called an address and the value is called the account. Each address points to an account where data is stored.​And everything on Solana is an address / account pair. For example, each program has an address and stores the code for the program at the associated account. (It&#x27;s actually a <a href="https://www.notion.so/Program-deploys-29780c48794c47308d5f138074dd9838" target="_blank" rel="nofollow noopener">bit more complex than this</a> but we don&#x27;t need to get into that...)​<h1>​Regular Accounts and Storage Accounts</h1>​Now let’s talk about addresses. On Solana an address is a 256 bit number that is most often encoded with base58, so it looks something like this <code>CXk5NCtYva7h4BcGXg5BDBDtZDwgBuWZcwmWt5ReY1yA</code>.​There are two different types of accounts — let’s call them Regular accounts and Storage accounts.​<h2>​Regular Accounts</h2>​Regular accounts are what you think of when you think of crypto addresses.​Bitcoin, Ethereum, and Solana all use Elliptic Curve cryptography which allows a public private keypair to sign messages. If you have a public private keypair, you can sign a message with your private key and other people can verify the message is signed correctly.​This public / private keypair mechanism allows wallets to sign transactions that authorize some action. The Bitcoin code won’t execute a transaction that sends Bitcoin unless that account has signed the transaction.​On Solana, the regular account corresponds to the public key. You hold the private key securely in your wallet and you can share your public key.​In this system, a public key is a 256 bit number which lies <a href="https://www.youtube.com/watch?v=NF1pwjL9-DE&amp;ab_channel=Computerphile" target="_blank" rel="nofollow noopener">on an elliptic curve</a>. So a valid Solana Regular Account is one which lies on the elliptic curve.​Regular accounts are used for a couple things:<ul><li>​Wallets — If you want to have an account that stores SOL, you will create a regular account. SOL is stored directly in your wallet. (Tokens / NFTs are stored on Storage Accounts associated with your wallet.)</li><li>​Programs — Every program is a regular account. It stores data in the account that is executed when the program is called.</li></ul>​<h2>​Storage Accounts (PDAs)</h2>​Storage accounts allow you to store data associated with programs. A program can do anything from storing the state of an NFT listing, allowing an escrow between two parties, to acting as a domain registry.​You know an account is a storage account because its address is not on the elliptic curve that we mentioned above. That means a storage account address is not part of a public / private keypair, it’s just a number.​Regular accounts can sign messages with their private key to authorize transactions. Storage accounts can’t do this since they don’t have a private key. Instead storage accounts are linked to a controller program.​So if you are creating a program that will take a poll of where everyone wants to go on vacation, that program will store information about the different vacation destinations and votes at storage programs that are linked to the program.​<h2>​How NFTs Work</h2>​If you ask me, the NFT spec in Solana is as complicated as possible (and I really want to rewrite it...). You can’t actually understand the NFT spec without understanding the coin spec.​A coin (aka a <a href="https://spl.solana.com/token" target="_blank" rel="nofollow noopener">token</a>) on Solana has a few characteristics:<ul><li>​An ID — This is called the Mint and is stored at the Mint account. The Mint account is a Storage Account that is controlled by the Solana Token Program.</li><li>​Metadata — Name, symbol, price, etc. This information is stored in accounts on chain and places off chain.</li><li>​Total Supply — Let’s say there are 1 million VIC coin. This VIC could be stored on one account or spread out across a bunch of different accounts. Each account stores a certain number of VIC and points to the owner. So I could have multiple accounts that store VIC. My overall balance is the sum of the VIC in these accounts.</li></ul>​Now that we understand how coins work, we turn to NFTs. An NFT is just a coin with a supply of 1. So each NFT has the following accounts (which are all Storage Accounts):<ul><li>​Mint — This is the ID of the NFT and stores some basic metadata.</li><li>​Metadata — This has even more Metadata about the NFT. For historical reasons, it’s another storage account...</li><li>​Token — This stores who owns the NFT. This can be super confusing since the address of the token account can change over time.</li></ul>​So if you own an NFT, what that means is the Token account for the NFT says that you are the owner.</div>

The Solana Programming Model

Ethereum is Hard
On Ethereum, it's incredibly difficult to figure out what NFTs someone owns.
Each Ethereum NFT is linked to a smart contract that minted the NFT. But since every NFT collection deploys their own smart contract, there are millions (?) of smart contracts that you need to scan over and check. 
When implementing token gated events for Luma, we used the Opensea's API to query for the NFTs that an address owns. Opensea has already done the hard work for us by scanning over all of Ethereum's contracts and blocks. 
But this kinda sucks, doesn't it? The blockchain is supposed to be a decentralized database and we just ended up querying a centralized provider...
Solana is Different
On Solana, we store data across the blockchain in key / value pairs. The terminology is kind of confusing, but each key / value pair is called an account. And everything is an account... But let's not get too sidetracked. 
When you mint a Solana NFT, you have a few different accounts involved in the transaction:
your Wallet
the Solana Token Program — This is part of the Solana Program Library (the standard library for Solana) and is the program we are running to mint the NFT.
the Mint account — This account serves as the ID of the NFT and it owns the metadata about the NFT.
the Associated Token Account — This is generated by combining your Wallet address with the Mint. As a consequence, you will have at most one Associated Token Account for each type of token. 
The last account is the most interesting. That's where we will store that you own the NFT. 
So if wanted to check if wallet WALRUS owns an NFT with the mint PUNK, we would look up the Associated Token Account by combining WALRUS and PUNK with a special function, let's say that outputs WALPUNK. Then we would read the data in WALPUNK to see if WALRUS owns PUNK. 
Now we are at parity with Ethereum. Given an NFT and a wallet, we can see who owns it. But we can do so much more...
On Solana, each validator node stores the full state of the system. This state is stored in what Solana calls the Accounts DB. The Accounts DB lets us do queries that would be impossible on Ethereum. 
In this case, we can easily go from a Wallet address to a list of all of the Associated Token Programs associated with that Wallet. 
Tada! On Solana, we can quickly find all the NFTs (and other tokens) that someone owns. And we don't have to rely on a centralized service like Opensea. 

Appendix: Diving into the Technical Details
This covers how we go from a Wallet address to all of the Associated Token Accounts associated with the Wallet. From those accounts, it's relatively straightforward to find information on the NFTs someone owns (although, I can write about that if people are interested...). 
web3
Using Solana's web3 library, you call getParsedTokenAccountsByOwner.
This makes a getTokenAccountsByOwner  RPC request to the URL that you specified when creating the collection. 
Solana Runtime
This request gets processed by a node running the Solana Runtime. The runtime interprets the request and calls a Rust function get_token_accounts_by_owner. This in turn calls several functions: 
get_token_accounts_by_owner
get_filtered_spl_token_accounts_by_owner
get_filtered_indexed_accounts
accounts.load_by_index_key_with_filter
accounts_db.index_scan_accounts
accounts_index.index_scan_accounts
That last step is interesting! That's the Accounts DB that we mentioned above. Each node running Solana keeps the full Account DB in memory which makes it easy to query and find data. 

OK, so I'm sure that a lot of that went over your head. A lot of it is still pretty confusing to me and I'm sure that I'm getting some of the details wrong. So if you have questions, ping me on Twitter!
And if you want to build dope, meaningful stuff on top of Solana, join us!

<div class="mirror-content" data-reactroot=""><h1>​Ethereum is Hard</h1>​On Ethereum, it&#x27;s incredibly difficult to figure out what NFTs someone owns.​Each Ethereum NFT is <a href="https://eips.ethereum.org/EIPS/eip-721" target="_blank" rel="nofollow noopener">linked to a smart contract</a> that minted the NFT. But since every NFT collection deploys their own smart contract, there are millions (?) of smart contracts that you need to scan over and check. ​When implementing <a href="https://lu.ma/crypto" target="_blank" rel="nofollow noopener">token gated events</a> for Luma, we used the <a href="https://docs.opensea.io/reference/api-overview" target="_blank" rel="nofollow noopener">Opensea&#x27;s API</a> to query for the NFTs that an address owns. Opensea has already done the hard work for us by scanning over all of Ethereum&#x27;s contracts and blocks. ​But this kinda sucks, doesn&#x27;t it? The blockchain is supposed to be a decentralized database and we just ended up querying a centralized provider...<h1>​Solana is Different</h1>​On Solana, we store data across the blockchain in key / value pairs. The terminology is kind of confusing, but each key / value pair is called an account. And everything is an account... But let&#x27;s not get too sidetracked. ​When you mint a Solana NFT, you have a few different accounts involved in the transaction:<ul><li>​your Wallet</li><li>​the <a href="https://spl.solana.com/token" target="_blank" rel="nofollow noopener">Solana Token Program</a> — This is part of the Solana Program Library (the standard library for Solana) and is the program we are running to mint the NFT.</li><li>​the Mint account — This account serves as the ID of the NFT and it owns the metadata about the NFT.</li><li>​the <a href="https://spl.solana.com/associated-token-account" target="_blank" rel="nofollow noopener">Associated Token Account</a> — This is generated by combining your Wallet address with the Mint. As a consequence, you will have at most one Associated Token Account for each type of token. </li></ul>​The last account is the most interesting. That&#x27;s where we will store that you own the NFT. ​So if wanted to check if wallet <code>WALRUS</code> owns an NFT with the mint <code>PUNK</code>, we would look up the Associated Token Account by combining <code>WALRUS</code> and <code>PUNK</code> with a special function, let&#x27;s say that outputs <code>WALPUNK</code>. Then we would read the data in <code>WALPUNK</code> to see if <code>WALRUS</code> owns <code>PUNK</code>. ​Now we are at parity with Ethereum. Given an NFT and a wallet, we can see who owns it. But we can do so much more...​On Solana, each validator node stores the full state of the system. This state is stored in what Solana calls the Accounts DB. The Accounts DB lets us do queries that would be impossible on Ethereum. ​In this case, we can easily go from a Wallet address to a list of all of the Associated Token Programs associated with that Wallet. ​Tada! On Solana, we can quickly find all the NFTs (and other tokens) that someone owns. And we don&#x27;t have to rely on a centralized service like Opensea. <hr/><h1>​Appendix: Diving into the Technical Details</h1>​This covers how we go from a Wallet address to all of the Associated Token Accounts associated with the Wallet. From those accounts, it&#x27;s relatively straightforward to find information on the NFTs someone owns (although, I can write about that if people are interested...). <h2>​web3</h2>​Using Solana&#x27;s web3 library, you call <a href="https://github.com/solana-labs/solana-web3.js/blob/10548b82e08ba865c36e2d86974fe15bce3fc566/src/connection.ts#L2333" target="_blank" rel="nofollow noopener">getParsedTokenAccountsByOwner</a>.​This makes a <code>getTokenAccountsByOwner</code> RPC request to the URL that you specified when creating the collection. <h2>​Solana Runtime</h2>​This request gets processed by a node running the Solana Runtime. The runtime interprets the request and calls a <a href="https://github.com/solana-labs/solana/blob/807263596748508056ccfef6be31527661387f03/rpc/src/rpc.rs#L1685" target="_blank" rel="nofollow noopener">Rust function</a> <code>get_token_accounts_by_owner</code>. This in turn calls several functions: <ol><li>​<code>get_token_accounts_by_owner</code></li><li>​<code>get_filtered_spl_token_accounts_by_owner</code></li><li>​<code>get_filtered_indexed_accounts</code></li><li>​<code>accounts.load_by_index_key_with_filter</code></li><li>​<code>accounts_db.index_scan_accounts</code></li><li>​<code>accounts_index.index_scan_accounts</code></li></ol>​That last step is interesting! That&#x27;s the Accounts DB that we mentioned above. Each node running Solana keeps the full Account DB in memory which makes it easy to query and find data. ​​OK, so I&#x27;m sure that a lot of that went over your head. A lot of it is still pretty confusing to me and I&#x27;m sure that I&#x27;m getting some of the details wrong. So if you have questions, <a href="https://twitter.com/VictorPontis" target="_blank" rel="nofollow noopener">ping me on Twitter</a>!​And if you want to build dope, meaningful stuff on top of Solana, join us!​​<img alt="Image" class="image" src="https://cdn.lu.ma/cdn-cgi/image/format=auto,dpr=2,quality=85,width=800/editor-images/km/c1f088f8-fb45-4ed7-8963-5d110b2a10e1" style="display:block;max-width:100%;max-height:400px;margin:24px auto"/></div>

How to Find the Solana NFT’s Someone Owns

How NFTs Work on Ethereum
The first NFTs were created on Ethereum. At first it was a free for all and every new NFT or NFT project encoded data in a proprietary way. But over time, a standard started to emerge and was codified in the ERC-721 standard.
Today, if you want to create a new NFT on Ethereum, you'll deploy a new smart contract on the blockchain that adheres to the ERC-721 standard.
So let's say that we are creating a new NFT project — Lumens. We would write a contract for Lumens and then deploy it to the blockchain and get a contract ID, say 123luma123contract. 
That contract keeps track of who owns which Lumen NFTs and has the power to mint new NFTs and transfer existing NFTs. If you own Lumen #5 and want to transfer it to someone, you send a signed message to the contract telling the contract to update ownership of Lumen #5.
So now we have a bunch of Lumen NFTs that are all grouped under one contract. This gives us the idea of a collection of NFTs. On Ethereum, an NFT is identified by the contract name + the token number. So Lumen#5 could also be referred to as 123luma123contract/5.
In theory, this all sounds nice but in practice it's much more messy. Part of the reason for the messiness is that everyone is writing and deploying their own ERC-721 contract. That's led to a ton of quirks and many NFT projects don't even properly adhere to the ERC-721 standard. 
Solana Does Things Differently
Now Solana does a lot of things differently, for better and for worse. 
On Ethereum for every NFT collection, you need to create a new ERC-721 contract. While this gives you more control, it's also complicated and expensive. It's like creating a new website by starting with some sample code. 
On Solana, when you create a new token or NFT, you use the existing programs already deployed that are part of Solana's Program Library. You can think of the SPL as analogous to the Python standard library. Sure, you could write your own version of map and sum but it's much easier to just use the builtins. 
The main drawback with Solana's Token Program is that it doesn't have a concept of NFTs or NFT collections. 
To create an NFT on Solana, we are creating a token but then setting some special parameters on the token to make it effectively an NFT. Specifically, a Solana NFT is a token with supply == 1 and decimals == 0. It's a little awkward, but it works.
Creating a collection is much more difficult. There's not a standardized, easy way of grouping a set of NFTs into a collection. While on Ethereum, if you mint multiple NFTs from the same contract, you can see they all came from the same contract. 
On Solana, you can set a Collection Name for your NFT. But that field isn't verified. So if I create the Lumen NFT collection, you could come along and decide to create an NFT with Collection Name == Lumen and pretend that your NFT is part of my collection. 
Then, how are people selling collections of NFTs on Solana? They are using hacks like verifying that the creator address of NFTs or just storing the list of NFTs in a collection in a web2 database. 
So while it's a lot easier and cheaper to get create a Solana NFT, there are a lot of things you need to figure out in order to create a successful NFT project. 
In a follow up post, I'll cover how to find the NFTs that someone owns given their wallet address.

<div class="mirror-content" data-reactroot=""><h2>​How NFTs Work on Ethereum</h2>​The first NFTs were created on Ethereum. At first it was a free for all and every new NFT or NFT project encoded data in a proprietary way. But over time, a standard started to emerge and was codified in the <a href="https://eips.ethereum.org/EIPS/eip-721" target="_blank" rel="nofollow noopener">ERC-721</a> standard.​Today, if you want to create a new NFT on Ethereum, you&#x27;ll deploy a new smart contract on the blockchain that adheres to the ERC-721 standard.​So let&#x27;s say that we are creating a new NFT project — Lumens. We would write a contract for Lumens and then deploy it to the blockchain and get a contract ID, say <code>123luma123contract</code>. ​That contract keeps track of who owns which Lumen NFTs and has the power to mint new NFTs and transfer existing NFTs. If you own <code>Lumen #5</code> and want to transfer it to someone, you send a signed message to the contract telling the contract to update ownership of <code>Lumen #5</code>.​So now we have a bunch of Lumen NFTs that are all grouped under one contract. This gives us the idea of a collection of NFTs. On Ethereum, an NFT is identified by the contract name + the token number. So <code>Lumen#5</code> could also be referred to as <code>123luma123contract/5</code>.​In theory, this all sounds nice but in practice it&#x27;s much more messy. Part of the reason for the messiness is that everyone is writing and deploying their own ERC-721 contract. That&#x27;s led to a ton of quirks and many NFT projects don&#x27;t even properly adhere to the ERC-721 standard. <h2>​Solana Does Things Differently</h2>​Now Solana does a lot of things differently, for better and for worse. ​On Ethereum for every NFT collection, you need to create a new ERC-721 contract. While this gives you more control, it&#x27;s also complicated and expensive. It&#x27;s like creating a new website by starting with some sample code. ​On Solana, when you create a new token or NFT, you use the existing programs already deployed that are part of Solana&#x27;s Program Library. You can think of the SPL as analogous to the Python standard library. Sure, you could write your own version of <code>map</code> and <code>sum</code> but it&#x27;s much easier to just use the builtins. ​The main drawback with Solana&#x27;s Token Program is that it doesn&#x27;t have a concept of NFTs or NFT collections. ​To create an NFT on Solana, we are creating a token but then setting some special parameters on the token to make it effectively an NFT. Specifically, a Solana NFT is a token with <code>supply == 1</code> and <code>decimals == 0</code>. It&#x27;s a little awkward, but it works.​Creating a collection is much more difficult. There&#x27;s not a standardized, easy way of grouping a set of NFTs into a collection. While on Ethereum, if you mint multiple NFTs from the same contract, you can see they all came from the same contract. ​On Solana, you can set a <code>Collection Name</code> for your NFT. But that field isn&#x27;t verified. So if I create the Lumen NFT collection, you could come along and decide to create an NFT with <code>Collection Name == Lumen</code> and pretend that your NFT is part of my collection. ​Then, how are people selling collections of NFTs on Solana? They are using hacks like verifying that the creator address of NFTs or just storing the list of NFTs in a collection in a web2 database. ​So while it&#x27;s a lot easier and cheaper to get create a Solana NFT, there are a lot of things you need to figure out in order to create a successful NFT project. ​In a follow up post, I&#x27;ll cover how to find the NFTs that someone owns given their wallet address.</div>

How Solana NFTs Differ from Ethereum

Running Without Alcohol
I've had a couple running goals this year — to run 750 miles for the year and to break 5 minutes in the mile. 

For the first goal, I'm at 705 miles for the year so if I run 20 miles the next couple weeks I should hit my goal. 
Two months ago I attempted to run a 5 minute mile. It was a really awkwardly paced race where I went out way too fast and ended up running a 5:08.
After that mile, I decided that I was going to give up drinking until I ran a 5 minute mile. Today, two months later, I haven't yet run a 5 minute mile, but I've stuck to giving up drinking. 
As someone who drank frequently at dinner and parties, I expected that it would be hard to give up alcohol, that I would be pressured to drink by friends and by a sense of social awkwardness.
That hasn't happened. Giving up drinking has been remarkably smooth. I've still had fun at parties. Awkward social situations haven't been more awkward. 
It's actually been a great discovery to see how little alcohol is actually needed in my social life. 
But giving up drinking was supposed to make me 8 seconds faster and so far, it hasn't done it's job! 
So I'm going to give up my sobriety streak and refocus on getting my miles and track workouts in every week. 🍺🏃
[Luma] Editor and Email Improvements
I've spent a lot of time the last few weeks working on improving the email workflows in Luma.
We have a much nicer editor where you can add images, buttons, or videos inline. 
You can fully customize the subject and body of any event related email.
You can see who opened your emails and who may need an extra reminder. 
You can schedule newsletters.
Aloha Dinner
On Friday, Calvin and I got dinner with a friend's parents. We didn't coordinate, but we showed up wearing nearly matching shirts 💐.

Coming Up
This week I'll be:
Learning Solana for an upcoming Luma product
Welcoming my mom to NYC who is visiting
Closing the gap on 750 running miles for the year

<div class="mirror-content" data-reactroot=""><h2>​Running Without Alcohol</h2>​I&#x27;ve had a couple running goals this year — to run 750 miles for the year and to break 5 minutes in the mile. <img alt="Image" class="image" src="https://cdn.lu.ma/cdn-cgi/image/format=auto,dpr=2,quality=85,width=415/editor-images/ne/7bf00539-9874-4482-bf26-d0a129a3303b" style="display:block;max-width:100%;max-height:100%;margin:24px auto"/>​For the first goal, I&#x27;m at 705 miles for the year so if I run 20 miles the next couple weeks I should hit my goal. ​Two months ago I attempted to run a 5 minute mile. It was a really awkwardly paced race where I went out way too fast and ended up running a 5:08.​After that mile, I decided that I was going to give up drinking until I ran a 5 minute mile. Today, two months later, I haven&#x27;t yet run a 5 minute mile, but I&#x27;ve stuck to giving up drinking. ​As someone who drank frequently at dinner and parties, I expected that it would be hard to give up alcohol, that I would be pressured to drink by friends and by a sense of social awkwardness.​That hasn&#x27;t happened. Giving up drinking has been remarkably smooth. I&#x27;ve still had fun at parties. Awkward social situations haven&#x27;t been more awkward. ​It&#x27;s actually been a great discovery to see how little alcohol is actually needed in my social life. ​But giving up drinking was supposed to make me 8 seconds faster and so far, it hasn&#x27;t done it&#x27;s job! ​So I&#x27;m going to give up my sobriety streak and refocus on getting my miles and track workouts in every week. 🍺🏃<h2>​[Luma] Editor and Email Improvements</h2>​I&#x27;ve spent a lot of time the last few weeks working on <a href="https://lu.ma/p/hqbp5cVc21gBpaE/Email-Improvements-in-Luma" target="_blank" rel="nofollow noopener">improving the email workflows in Luma</a>.<ul><li>​We have a much nicer editor where you can add images, buttons, or videos inline. </li><li>​You can fully customize the subject and body of any event related email.</li><li>​You can see who opened your emails and who may need an extra reminder. </li><li>​You can schedule newsletters.</li></ul><h2>​Aloha Dinner</h2>​On Friday, <a href="https://lu.ma/calvinfo" target="_blank" rel="nofollow noopener">Calvin</a> and I got dinner with <a href="https://lu.ma/g" target="_blank" rel="nofollow noopener">a friend&#x27;s</a> parents. We didn&#x27;t coordinate, but we showed up wearing nearly matching shirts 💐.<img alt="Image" class="image" src="https://cdn.lu.ma/cdn-cgi/image/format=auto,dpr=2,quality=85,width=459/editor-images/1w/e7fcb103-2615-42e5-9fa7-4e83d14980e3" style="display:block;max-width:100%;max-height:100%;margin:24px auto"/><h2>​Coming Up</h2>​This week I&#x27;ll be:<ul><li>​Learning <a href="https://solana.com/" target="_blank" rel="nofollow noopener">Solana</a> for an upcoming Luma product</li><li>​Welcoming my mom to NYC who is visiting</li><li>​Closing the gap on 750 running miles for the year</li></ul></div>

Week Update — December 12th

We've significantly revamped the email experience in Luma across editing, events, and newsletters! 
Improved Editing Experience
We have made major improvements to our editor so that you can focus on what's important — your writing.
We have a new way of adding blocks and have improved the look and feel of the editor. Check it out below or play with here.


Event Email Customization
For every Luma event, we automatically send registration and reminder emails. You can now customize the subject and body of those emails. 
On your manage page, check out the Emails tab and start personalizing your emails 😇.

Track Email Opens
We keep track of email statistics for every email you send. That way it's easy to see who opened your email and who may need a little extra reminder...

Schedule Newsletters
Newsletters are a great way to promote upcoming events and stay in touch with your community. You can import contacts, tag your audience, and follow up with event guests. 
Today, you can now schedule your newsletters! When sending a newsletter, just choose Send Later and then pick a time.

<div class="mirror-content" data-reactroot="">​We&#x27;ve significantly revamped the email experience in Luma across editing, events, and newsletters! <h2>​Improved Editing Experience</h2>​We have made major improvements to our editor so that you can focus on what&#x27;s important — your writing.​We have a new way of adding blocks and have improved the look and feel of the editor. Check it out below or <a href="http://lu.ma/home/newsletters" target="_blank" rel="nofollow noopener">play with here</a>.<img alt="Image" class="image" src="https://cdn.lu.ma/editor-images/nf/b7c4522e-e943-4310-8a2a-81397b07a2e6.gif" style="display:block;max-width:100%;max-height:400px;margin:24px auto"/>​<h2>​Event Email Customization</h2>​For every Luma event, we automatically send registration and reminder emails. You can now customize the subject and body of those emails. ​On your manage page, check out the Emails tab and start personalizing your emails 😇.<img alt="Image" class="image" src="https://cdn.lu.ma/editor-images/ka/2d768272-ab49-44d7-babd-fd0c63e6c172.gif" style="display:block;max-width:100%;max-height:100%;margin:24px auto"/><h2>​Track Email Opens</h2>​We keep track of email statistics for every email you send. That way it&#x27;s easy to see who opened your email and who may need a little extra reminder...<img alt="Image" class="image" src="https://cdn.lu.ma/cdn-cgi/image/format=auto,dpr=2,quality=85,width=414/editor-images/o0/900fad1b-9820-417b-9ce6-7abc5169abb3" style="display:block;max-width:100%;max-height:100%;margin:24px auto"/><h2>​Schedule Newsletters</h2>​Newsletters are a great way to promote upcoming events and stay in touch with your community. You can import contacts, tag your audience, and follow up with event guests. ​Today, you can now schedule your newsletters! When sending a newsletter, just choose Send Later and then pick a time.<img alt="Image" class="image" src="https://cdn.lu.ma/cdn-cgi/image/format=auto,dpr=2,quality=85,width=429/editor-images/l8/ab5abaea-7046-4fe0-ae41-afcae03e6d1a" style="display:block;max-width:100%;max-height:100%;margin:24px auto"/></div>

Email Improvements in Luma

So I've dropped off writing these, but I'm back! 
For a refresher, I write a weekly email on Sundays that covers what I've been up to along with goals. I find it helpful for both reflection and planning. All that to say... hi, again!
Thanksgiving in Solana Beach

For Thanksgiving, I flew to San Diego and spent a week with my mom and her dog staying at an AirBnB near the beach. I haven't minded the winter in New York, but it was a nice change of pace. 
On Black Friday I hosted my high school's 10 year reunion at a nearby brewery. I didn't know what to expect but it was ridiculously fun. If you have a big reunion coming up, I recommend you go! Everyone was so much nicer than high school and most people I hadn't seen since leaving San Diego. 
Dune
I saw Dune in IMAX the day it came out with friends who had read the books and built up some friendgroup hype. I loved Dune — the story, the visuals, the expansiveness. It reminded me of first watching Lord of the Rings when I was 8 or 9 when I saw each movie more than 3 times.
So now I'm listening to the Dune audio book which is also incredible. It's more cinematic than any audiobook I've listened to replete with sound effects, music, and voice actors. 
Luma — Emails and Solana
For Luma, I've been working on some big improvements to emails within Luma. You're now able to customize the registration and reminder emails that go out with each event. And there are a few more fun things coming up this week...
I've also been playing around with Solana which we are going to incorporate into lu.ma/crypto. This week I plan on releasing a public Solana API along with a short writeup. 

P.S. I'm finishing up this email just in time to start the Advent of Code problem tonight. I'm doing it in Rust and you can find my solutions here.
Until next week!

<div class="mirror-content" data-reactroot="">​So I&#x27;ve dropped off writing these, but I&#x27;m back! ​For a refresher, I write a weekly email on Sundays that covers what I&#x27;ve been up to along with goals. I find it helpful for both reflection and planning. All that to say... hi, again!<h2>​Thanksgiving in Solana Beach</h2><img alt="Image" class="image" src="https://cdn.lu.ma/cdn-cgi/image/format=auto,dpr=2,quality=85,width=372/editor-images/nn/c8bf0442-0130-49be-b6d0-630351a65cd3" style="display:block;max-width:100%;max-height:100%;margin:24px auto"/>​For Thanksgiving, I flew to San Diego and spent a week with my mom and her dog staying at an AirBnB near the beach. I haven&#x27;t minded the winter in New York, but it was a nice change of pace. ​On Black Friday I hosted my high school&#x27;s 10 year reunion at a nearby brewery. I didn&#x27;t know what to expect but it was ridiculously fun. If you have a big reunion coming up, I recommend you go! Everyone was so much nicer than high school and most people I hadn&#x27;t seen since leaving San Diego. <h2>​Dune</h2>​I saw Dune in IMAX the day it came out with friends who had read the books and built up some friendgroup hype. I loved Dune — the story, the visuals, the expansiveness. It reminded me of first watching Lord of the Rings when I was 8 or 9 when I saw each movie more than 3 times.​So now I&#x27;m listening to the <a href="https://www.amazon.com/Dune-Frank-Herbert-audiobook/dp/B000R34YKC" target="_blank" rel="nofollow noopener">Dune audio book</a> which is also incredible. It&#x27;s more cinematic than any audiobook I&#x27;ve listened to replete with sound effects, music, and voice actors. <h2>​Luma — Emails and Solana</h2>​For Luma, I&#x27;ve been working on some big improvements to emails within Luma. You&#x27;re now able to customize the registration and reminder emails that go out with each event. And there are a few more fun things coming up this week...​I&#x27;ve also been playing around with <a href="https://solana.com/" target="_blank" rel="nofollow noopener">Solana</a> which we are going to incorporate into <a href="http://lu.ma/crypto." target="_blank" rel="nofollow noopener">lu.ma/crypto.</a> This week I plan on releasing a public Solana API along with a short writeup. ​​P.S. I&#x27;m finishing up this email just in time to start the <a href="https://adventofcode.com/" target="_blank" rel="nofollow noopener">Advent of Code</a> problem tonight. I&#x27;m doing it in Rust and you can <a href="https://github.com/vpontis/advent-rust" target="_blank" rel="nofollow noopener">find my solutions here</a>.​Until next week!</div>

Week Update — Dec 5, 2021

I'm writing this here since I'm going to want to do it some point in the future and this info is incredibly hard to find. 
If you back up your iPhone to your Mac, you will find in that backup an unencrypted SQLite file that is your message DB.
The iOS backup is located at: /Users/victor/Library/Application Support/MobileSync/Backup/
Within the folder of your specific backup, you can find the message file here: 3d/3d3d0d7e5fb2ce288813306e4d4636395e047a3d28
The most interesting table is the message table and to convert a date to a reasonable value you can do:
DateTime.from('2001-01-01').plus({seconds: date / 1_000_000_000})

<div class="mirror-content" data-reactroot="">​I&#x27;m writing this here since I&#x27;m going to want to do it some point in the future and this info is incredibly hard to find. ​If you back up your iPhone to your Mac, you will find in that backup an unencrypted SQLite file that is your message DB.​The iOS backup is located at: <code>/Users/victor/Library/Application Support/MobileSync/Backup/</code>​Within the folder of your specific backup, you can find the message file here: <code>3d/3d3d0d7e5fb2ce288813306e4d4636395e047a3d28</code>​The most interesting table is the <code>message</code> table and to convert a date to a reasonable value you can do:<pre>​DateTime.from(&#x27;2001-01-01&#x27;).plus({seconds: date / 1_000_000_000})</pre></div>

How to Extract iOS Messages from your iOS Backup

Overview
When we were creating email templates with Luma, we consistently ran into two issues:
Email has weird CSS. We can't use the same CSS that you use in the browser because email clients support only a weird subset of CSS rules . We need some way of creating a button component that actually looks like a button. 
We want to use React. React allows us to reuse components across emails, embed logic into the views, and render email previews in the browser. 
Unfortunately, we couldn't find an existing solution that satisfactorily fixed these two issues.
Email providers like Postmark, Sendgrid, and Mailjet give you some tools to create email templates but they are hard to embed logic into and these proprietary templates will lock you into one email provider. 
The React ecosystem has a few packages for creating emails in React, but none of these packages are popular and they don't give you built in components that handle email CSS. 
Our Solution
To solve the two issues, we built a system that allows us to couple React with MJML which is an open source email template language built by Mailjet. MJML gives you components like mj-button which gives you a reasonable looking button by creating nested <table /> s.
We built a library @luma-team/mjml-react which gives you React components that wrap the MJML templating language. So you can just write <MjmlButton /> and you'll get a reasonable looking button. 
Moving our templating logic to React has also allowed us to reuse logic across templates, easily deploy new templates, and gives us freedom to switch email providers in the future. 
Preview Emails in the Browser
We made sure that we can preview these email templates in the browser. This makes it easy to iterate on email templates and lets us build a template gallery that anyone on the team can look through.
You can see our email template library here: lu.ma/style-guide/email
Example Repo
If you want to build emails like we do, I've set up an example repo that shows how you can use MJML to build templates, preview them in the browser, and then send them with Postmark. 
You can play with it and even send some test emails 😍.
Preview:  https://email-templates-starter-repo.luma-dev.com/ 
 GitHub: https://github.com/luma-team/email-templates-starter-repo


P.S. If you think creating clean solutions to common engineering problems is fun, you should join us at Luma !

<div class="mirror-content" data-reactroot=""><h2>​Overview</h2>​When we were creating email templates with Luma, we consistently ran into two issues:<ol><li>​Email has weird CSS. We can&#x27;t use the same CSS that you use in the browser because <a href="https://www.campaignmonitor.com/css/" target="_blank" rel="nofollow noopener">email clients support only a weird subset of CSS rules</a>. We need some way of creating a button component that actually looks like a button. </li><li>​We want to use React. React allows us to reuse components across emails, embed logic into the views, and render email previews in the browser. </li></ol>​Unfortunately, we couldn&#x27;t find an existing solution that satisfactorily fixed these two issues.​Email providers like Postmark, Sendgrid, and Mailjet give you some tools to create email templates but they are hard to embed logic into and these proprietary templates will lock you into one email provider. ​The React ecosystem has a few packages for creating emails in React, but none of these packages are popular and they don&#x27;t give you built in components that handle email CSS. <h2>​Our Solution</h2>​To solve the two issues, we built a system that allows us to couple React with <a href="https://mjml.io/" target="_blank" rel="nofollow noopener">MJML</a> which is an open source email template language built by Mailjet. MJML gives you components like <code>mj-button</code> which gives you a reasonable looking button by creating nested <code>&lt;table /&gt;</code>s.​We built a library <a href="https://github.com/luma-team/mjml-react" target="_blank" rel="nofollow noopener">@luma-team/mjml-react</a> which gives you React components that wrap the MJML templating language. So you can just write &lt;MjmlButton /&gt; and you&#x27;ll get a reasonable looking button. ​Moving our templating logic to React has also allowed us to reuse logic across templates, easily deploy new templates, and gives us freedom to switch email providers in the future. ​Preview Emails in the Browser​We made sure that we can preview these email templates in the browser. This makes it easy to iterate on email templates and lets us build a template gallery that anyone on the team can look through.​You can see our email template library here: <a href="http://lu.ma/style-guide/email" target="_blank" rel="nofollow noopener">lu.ma/style-guide/email</a>​Example Repo​If you want to build emails like we do, I&#x27;ve set up an example repo that shows how you can use MJML to build templates, preview them in the browser, and then send them with Postmark. ​You can play with it and even send some test emails 😍.​Preview: <a href="https://email-templates-starter-repo.luma-dev.com/" target="_blank" rel="nofollow noopener">https://email-templates-starter-repo.luma-dev.com/</a> GitHub: <a href="https://github.com/luma-team/email-templates-starter-repo" target="_blank" rel="nofollow noopener">https://github.com/luma-team/email-templates-starter-repo</a><img class="image" src="https://cdn.lu.ma/cdn-cgi/image/format=auto,dpr=2,quality=85,width=800/editor-images/ms/cadc16be-9a8b-4859-8dce-4a538d20a0e3" style="display:block;max-width:100%;max-height:400px;margin:24px auto"/><hr/>​P.S. If you think creating clean solutions to common engineering problems is fun, you should <a href="http://lu.ma/jobs" target="_blank" rel="nofollow noopener">join us at Luma</a>!</div>

Write Beautiful Emails in React

Demo — GitHub — Thanks to Stanislaw Chmiela for working on this with me.
While the hierarchical React state model works well most of the time, you can run into situations where the hierarchical structure gives your website very bad performance. This can often come up if you are changing state that affects deep leaf nodes since that can cause a re-render of the full React tree.
For example, we ran into performance issues when rendering presence status on a Luma Community page.
We render a green dot on someone’s avatar if they are currently in the Luma Community. After they’ve been inactive for 30 seconds, that avatar turns orange.
For Luma, it’s very important that presence is very real time because if you message someone who is online, you should expect they will see your message. That means in a large community we may need to re-render presence information every second.


Performance Issues
The traditional way of hosting presence state with React is as follows:
const CommunityHome = () => {
 const presenceInfos = usePresence();

 return (
 <CommunityHomeContent>
 <MainContent />
 <Sidebar>
 {presenceInfos.map((presence) => 
 <MemberRow key={presence.id} presence={presence} />
 )}
 </Sidebar>
 </CommunityHomeContent>
 )
}
Or here is a way that you can use React Context to avoid prop drilling:
const CommunityHome = () => {
 return (
 <CommunityHomeContent>
 <MainContent />
 <Sidebar>
 {members.map((member) => (
 <MemberRow key={member.id} member={member} />
 )}
 </Sidebar>
 </CommunityHomeContent>
 )
}

const MemberRow = ({member}) => {
 const {presenceInfos} = useContext(PresenceContext);
 const presenceInfo = presenceInfos.find((presence) => presence.member_id = member.id);

 return <div>...</div>;
}
But both of these have the same performance issue. Whenever one presenceInfo changes we have to re-render the whole page and all of the <MemberRow /> s.

Introducing Tangle
So to solve this problem, we built Tangle. Tangle is a lightweight state management system that uses event listeners to only re-render affected leaf components.
This system is super simple, performant, and can be easily dropped into your existing codebase.
Here's an example of the results we can get with Tangle:


Usage
type HighlightEvent = {
 type: `highlight-${number}`;
 idx: number;
 color: typeof COLORS[number];
};
const HighlightStore = new TangleStore<HighlightEvent>();

const Home = () => {
 useEffect(() => {
 setInterval(() => {
 const randomIndex = Math.floor(Math.random() * NUM_CELLS);

 HighlightStore.dispatchEvent({
 type: `highlight-${randomIndex}`,
 idx: randomIndex,
 color: _.sample(COLORS) as typeof COLORS[number],
 });

 setTimeout(() => {
 HighlightStore.dispatchEvent({
 type: `highlight-${randomIndex}`,
 idx: randomIndex,
 color: "gray",
 });
 }, 600);
 }, 10);
 }, []);
 
 return (
 <Container>
 <Input />
 <Grid />
 </Container>
 );
}

// This is the leaf component
const Cell = ({ idx }: { idx: number }) => {
 const [color, setColor] = useState("gray");

 useTangle({
 store: HighlightStore,
 eventName: `highlight-${idx}`,
 handleEvent: (event) => {
 setColor(event.color);
 },
 });

 return (
 <div key={idx} className={"cell"} style={{backgroundColor: color}} />
 );
};
If you want to use Tangle, you can use the code from the Github Repo and copy it into your codebase. Reach out to us if you start using it and we can add the code to npm 🙂.

Existing Solutions
React has had this problem with deeply nested state since its release, so there are a lot of existing solutions.
Redux
Redux solves this by using connect() which allows a leaf component to listen to a selection of state.
MobX
MobX introduces an event driven model to state and provides decorators that allow you to listen to a specific state value.
Recoil
Recoil is a new library from Facebook that breaks down your state into different atoms. It is powerful and allows you to compose different atoms.
useContextSelector
This allows you to create a selector on a context to only re-render a component when that selector changes. React is considering bringing in this function into native React.
Unfortunately this will still require every component to run a check when the context changes before they can bail out of rendering.
Why these don’t work for us
While all these libraries are great, we felt they add significant educational overhead. They aren’t designed to be dropped into a project that is already using a different state management system.

<div class="mirror-content" data-reactroot="">​<a href="https://tangle-demo.luma-dev.com/" target="_blank" rel="nofollow noopener">Demo</a> — <a href="https://github.com/luma-team/tangle-demo" target="_blank" rel="nofollow noopener">GitHub</a> — Thanks to <a href="https://twitter.com/sjchmiela?lang=en" target="_blank" rel="nofollow noopener">Stanislaw Chmiela</a> for working on this with me.​While the hierarchical React state model works well most of the time, you can run into situations where the hierarchical structure gives your website very bad performance. This can often come up if you are changing state that affects deep leaf nodes since that can cause a re-render of the full React tree.​For example, we ran into performance issues when rendering presence status on a <a href="http://lu.ma/communities" target="_blank" rel="nofollow noopener">Luma Community</a> page.​We render a green dot on someone’s avatar if they are currently in the Luma Community. After they’ve been inactive for 30 seconds, that avatar turns orange.​For Luma, it’s very important that presence is very real time because if you message someone who is online, you should expect they will see your message. That means in a large community we may need to re-render presence information every second.<img width="300" src="https://cdn.lu.ma/cdn-cgi/image/format=auto,dpr=2,quality=85,width=300/editor-images/0i/cfff18b3-ce7e-410b-8332-899ad1bbe1e6" style="display:block;max-width:100%;max-height:100%;margin:24px auto"/>​<h2>​Performance Issues</h2>​The traditional way of hosting presence state with React is as follows:<pre>​const CommunityHome = () =&gt; {
 const presenceInfos = usePresence();

 return (
 &lt;CommunityHomeContent&gt;
 &lt;MainContent /&gt;
 &lt;Sidebar&gt;
 {presenceInfos.map((presence) =&gt; 
 &lt;MemberRow key={presence.id} presence={presence} /&gt;
 )}
 &lt;/Sidebar&gt;
 &lt;/CommunityHomeContent&gt;
 )
}</pre>​Or here is a way that you can use React Context to avoid prop drilling:<pre>​const CommunityHome = () =&gt; {
 return (
 &lt;CommunityHomeContent&gt;
 &lt;MainContent /&gt;
 &lt;Sidebar&gt;
 {members.map((member) =&gt; (
 &lt;MemberRow key={member.id} member={member} /&gt;
 )}
 &lt;/Sidebar&gt;
 &lt;/CommunityHomeContent&gt;
 )
}

const MemberRow = ({member}) =&gt; {
 const {presenceInfos} = useContext(PresenceContext);
 const presenceInfo = presenceInfos.find((presence) =&gt; presence.member_id = member.id);

 return &lt;div&gt;...&lt;/div&gt;;
}</pre>​But both of these have the same performance issue. Whenever one <code>presenceInfo</code> changes we have to re-render the whole page and all of the <code>&lt;MemberRow /&gt;</code>s.​<h2>​Introducing Tangle</h2>​So to solve this problem, we built Tangle. Tangle is a lightweight state management system that uses event listeners to only re-render affected leaf components.​This system is super simple, performant, and can be easily dropped into your existing codebase.​Here&#x27;s an example of the results we can get with Tangle:<img src="https://cdn.lu.ma/editor-images/la/0a33f8eb-db3a-4ce8-96e1-cb2e89ba3ae6" style="display:block;max-width:100%;max-height:400px;margin:24px auto"/>​​Usage<pre>​type HighlightEvent = {
 type: `highlight-${number}`;
 idx: number;
 color: typeof COLORS[number];
};
const HighlightStore = new TangleStore&lt;HighlightEvent&gt;();

const Home = () =&gt; {
 useEffect(() =&gt; {
 setInterval(() =&gt; {
 const randomIndex = Math.floor(Math.random() * NUM_CELLS);

 HighlightStore.dispatchEvent({
 type: `highlight-${randomIndex}`,
 idx: randomIndex,
 color: _.sample(COLORS) as typeof COLORS[number],
 });

 setTimeout(() =&gt; {
 HighlightStore.dispatchEvent({
 type: `highlight-${randomIndex}`,
 idx: randomIndex,
 color: &quot;gray&quot;,
 });
 }, 600);
 }, 10);
 }, []);
 
 return (
 &lt;Container&gt;
 &lt;Input /&gt;
 &lt;Grid /&gt;
 &lt;/Container&gt;
 );
}

// This is the leaf component
const Cell = ({ idx }: { idx: number }) =&gt; {
 const [color, setColor] = useState(&quot;gray&quot;);

 useTangle({
 store: HighlightStore,
 eventName: `highlight-${idx}`,
 handleEvent: (event) =&gt; {
 setColor(event.color);
 },
 });

 return (
 &lt;div key={idx} className={&quot;cell&quot;} style={{backgroundColor: color}} /&gt;
 );
};</pre>​If you want to use Tangle, you can use the code from the <a href="https://github.com/luma-team/tangle-demo" target="_blank" rel="nofollow noopener">Github Repo</a> and copy it into your codebase. Reach out to us if you start using it and we can add the code to npm 🙂.​<h2>​Existing Solutions</h2>​React has had this problem with deeply nested state since its release, so there are a lot of existing solutions.​<a href="https://react-redux.js.org/" target="_blank" rel="nofollow noopener">Redux</a>​Redux solves this by using <a href="https://react-redux.js.org/api/connect" target="_blank" rel="nofollow noopener">connect()</a> which allows a leaf component to listen to a selection of state.​<a href="https://mobx.js.org/README.html" target="_blank" rel="nofollow noopener">MobX</a>​MobX introduces an event driven model to state and provides decorators that allow you to listen to a specific state value.​<a href="https://recoiljs.org/" target="_blank" rel="nofollow noopener">Recoil</a>​Recoil is a new library from Facebook that breaks down your state into different atoms. It is powerful and allows you to compose different atoms.​<a href="https://github.com/dai-shi/use-context-selector" target="_blank" rel="nofollow noopener">useContextSelector</a>​This allows you to create a selector on a context to only re-render a component when that selector changes. <a href="https://github.com/facebook/react/pull/20646" target="_blank" rel="nofollow noopener">React is considering</a> bringing in this function into native React.​Unfortunately this will still require every component to run a check when the context changes before they can bail out of rendering.​Why these don’t work for us​While all these libraries are great, we felt they add significant educational overhead. They aren’t designed to be dropped into a project that is already using a different state management system.</div>

Tangle — Event Driven State in React

This is the first project we are supporting as part of the  Luma Open Source Fellowship .
We chose to support  tiptap  because it's been an essential part of building out our text editor at Luma. There are so many editor libraries out there but we've appreciated how tiptap is incredibly well designed, extensible, and always improving. 
Fun fact: this post was written using the Luma Editor that was built on tiptap. 
The Makers — Philipp and Hans
Philipp Kühn: Creator of tiptap 
 Hans Pagel: Co-Creator & Maintainer of tiptap
Philipp and Hans both studied design, but ended up in development. That’s why they approach every project like a product, even the open source package tiptap which they started in 2018. Back then, all text editors came with ready-to-use styling, but they wanted to have  full control  about everything. 
That’s why they build their own rich text editor, which has a ton of extensions to add functionality, and a powerful API to write your own extensions, but comes  without any styling at all . It’s up to you! Use the framework you like, write the CSS that fits your workflow, create the editor experience you want. They will just back you up with the technology to build cool things.
By now, tiptap is one of the most popular text editors (~800,000 downloads/month) and used by teams like GitLab, Statamic or Nextcloud. With the backing of their own company, the digital product agency überdosis, and hundreds of sponsors from the community, they are able to work full-time on it and improve it every day.
The Event
We are hosting a conversation with Hans tomorrow at 11am Eastern / 5pm CEST.
Join the Conversation

P.S. We'd be remiss to support tiptap without acknowledging the great work of  ProseMirror  which tiptap is built on top of.

<div class="mirror-content" data-reactroot="">​This is the first project we are supporting as part of the <a href="https://lu.ma/p/r85D4lQFtv1QsCp/Introducing-the-Luma-Open-Source-Fellowship" target="_blank" rel="nofollow noopener">Luma Open Source Fellowship</a>.​We chose to support <a href="http://tiptap.dev" target="_blank" rel="nofollow noopener">tiptap</a> because it&#x27;s been an essential part of building out our text editor at Luma. There are so many editor libraries out there but we&#x27;ve appreciated how tiptap is incredibly well designed, extensible, and always improving. ​Fun fact: this post was written using the Luma Editor that was built on tiptap. <h2>​The Makers — Philipp and Hans</h2>​Philipp Kühn: Creator of tiptap Hans Pagel: Co-Creator &amp; Maintainer of tiptap​Philipp and Hans both studied design, but ended up in development. That’s why they approach every project like a product, even the open source package tiptap which they started in 2018. Back then, all text editors came with ready-to-use styling, but they wanted to have full control about everything. ​That’s why they build their own rich text editor, which has a ton of extensions to add functionality, and a powerful API to write your own extensions, but comes without any styling at all. It’s up to you! Use the framework you like, write the CSS that fits your workflow, create the editor experience you want. They will just back you up with the technology to build cool things.​By now, tiptap is one of the most popular text editors (~800,000 downloads/month) and used by teams like GitLab, Statamic or Nextcloud. With the backing of their own company, the digital product agency überdosis, and hundreds of sponsors from the community, they are able to work full-time on it and improve it every day.<h2>​The Event</h2>​We are hosting a conversation with Hans tomorrow at 11am Eastern / 5pm CEST.<div style="text-align:center;margin-bottom:1rem"><a href="https://lu.ma/tiptap" class="btn" style="background-color:#de3163;color:#ffffff;border:1px solid transparent;padding:0.75rem 1.5rem;border-radius:8px;font-size:1rem !important;line-height:1 !important;word-break:break-word;display:inline-block;text-align:center;font-weight:bold;text-decoration:none !important" target="_blank" rel="nofollow noopener">Join the Conversation</a></div>​​P.S. We&#x27;d be remiss to support tiptap without acknowledging the great work of <a href="https://prosemirror.net/" target="_blank" rel="nofollow noopener">ProseMirror</a> which tiptap is built on top of. ​<img width="414" src="https://cdn.lu.ma/cdn-cgi/image/format=auto,dpr=2,quality=85,width=414/editor-images/er/447b628c-32bd-4546-8db7-a546580185ea" style="display:block;max-width:100%;max-height:400px;margin:24px auto"/>​</div>

Open Source Fellowship: tiptap by Hans Pagel and Philipp Kühn

At Luma, like all software companies, we are building on top of the great work of the open source community. Luma wouldn't be where it is today without the tremendous work put in every day by developers across the world. 
So we want to give back. 
Today, we're launching the  Luma Open Source Fellowship  to support open source projects along with hardworking developers behind those projects.
Each month we'll pick one or two projects and they will receive:
A grant ranging from $1-5k
A blog post highlighting their project and work
A Luma event that will feature their work and will be uploaded to YouTube
Access to the Luma Open Source Fellowship community 
We're launching today with our support for  tiptap  made by Hans Pagel and Phillipp Kühn. We have a  post announcing our support for tiptap  and if you're available, you can  join our discussion with Phillip of tiptap  tomorrow at 11am Eastern.
If you'd like to nominate yourself or someone else for the fellowship, shoot us an email at  osf@lu.ma .

<div class="mirror-content" data-reactroot="">​At Luma, like all software companies, we are building on top of the great work of the open source community. Luma wouldn&#x27;t be where it is today without the tremendous work put in every day by developers across the world. ​So we want to give back. ​Today, we&#x27;re launching the Luma Open Source Fellowship to support open source projects along with hardworking developers behind those projects.​Each month we&#x27;ll pick one or two projects and they will receive:<ul><li>​A grant ranging from $1-5k</li><li>​A blog post highlighting their project and work</li><li>​A Luma event that will feature their work and will be uploaded to YouTube</li><li>​Access to the Luma Open Source Fellowship community </li></ul>​We&#x27;re launching today with our support for <a href="https://www.tiptap.dev/" target="_blank" rel="nofollow noopener">tiptap</a> made by Hans Pagel and Phillipp Kühn. We have a <a href="https://lu.ma/p/JgXb5d0FIGJIDZa/Open-Source-Fellowship-tiptap-by-Hans-Pagel-and-Phillipp-Khn" target="_blank" rel="nofollow noopener">post announcing our support for tiptap</a> and if you&#x27;re available, you can <a href="http://lu.ma/tiptap" target="_blank" rel="nofollow noopener">join our discussion with Phillip of tiptap</a> tomorrow at 11am Eastern.​If you&#x27;d like to nominate yourself or someone else for the fellowship, shoot us an email at <a href="http://osf@lu.ma" target="_blank" rel="nofollow noopener">osf@lu.ma</a>.​<img width="414" src="https://cdn.lu.ma/cdn-cgi/image/format=auto,dpr=2,quality=85,width=414/editor-images/dl/232123ab-a727-4fb3-a56e-adfa65d2cd52"/>​</div>

Introducing the Luma Open Source Fellowship

This is the first release of the Luma API. This is API is used by the Luma web and mobile clients so it's tested on production every day. 
But since this is just a beta release of the API, it is subject to breaking changes without announcement. Use at your own risk. 
Authentication
Create API Key
You can create an API Key that authenticates request on behalf of your account. Create your API Key here:  https://lu.ma/home/settings/integrations
An API Key with Account permissions lets someone take any action on the Luma API as you (including undocumented routes such as deleting your account). So be careful and do not share this API token. 
Authenticate Requests
You can authenticate requests by including your Account API Key in the Header field:  x-luma-api-key
API URLs + Format
The base of our API is:  https://api.lu.ma
The Luma API expects JSON as input and output. 
Example Request
The following request is a great way of testing that your API key is working and that your requests are configured properly. 
curl -H "x-luma-api-key: API_KEY_SECRET" https://api.lu.ma/user
Routes
POST /event/create
Request Body
name  — Required. The name of the event.
start_at  — The start of the event in a Date ISO string.
cover_url  — Valid URL for the event cover photo. Restricted to Unsplash or an image on the Luma CDN. 
duration_minutes  — The length of your event in minutes.
require_rsvp  — Require guests to register in order to see meeting information.
require_rsvp_approval  — Require host approval for a guest to see meeting information. If true,  require_rsvp  must also be true.
zoom_creation_method  —  "created-automatically" | "existing-attached" | "manually-entered" . If you want us to create a Zoom on your account, choose  "created-automatically" .
zoom_meeting_id
zoom_meeting_password
zoom_meeting_url
zoom_session_type  —  "meeting" | "webinar"
community_api_id  — The community that the event should be linked to. 
Response Object
api_id  — The ID of the event.
GET /user
Returns the current user if authenticated. If not authenticated, returns a 401.
POST /user/audience/import
Import members into your audience. Tag existing members.
Request Body
emails  — Required. An array of emails.
member_tag_api_ids  — Optional. An array of tag IDs to apply to the members.
Response Object
This route returns an empty object on success.

<div class="mirror-content" data-reactroot="">​This is the first release of the Luma API. This is API is used by the Luma web and mobile clients so it&#x27;s tested on production every day. ​But since this is just a beta release of the API, it is subject to breaking changes without announcement. Use at your own risk. <h1>​Authentication</h1><h2>​Create API Key</h2>​You can create an API Key that authenticates request on behalf of your account. Create your API Key here: <a href="https://lu.ma/home/settings/integrations" target="_blank" rel="nofollow noopener">https://lu.ma/home/settings/integrations</a>​An API Key with Account permissions lets someone take any action on the Luma API as you (including undocumented routes such as deleting your account). So be careful and do not share this API token. <h2>​Authenticate Requests</h2>​You can authenticate requests by including your Account API Key in the Header field: <code>x-luma-api-key</code><h2>​API URLs + Format</h2>​The base of our API is: <code>https://api.lu.ma</code>​The Luma API expects JSON as input and output. <h2>​Example Request</h2>​The following request is a great way of testing that your API key is working and that your requests are configured properly. <pre>​curl -H &quot;x-luma-api-key: API_KEY_SECRET&quot; https://api.lu.ma/user</pre><h1>​Routes</h1><h2>​<code>POST /event/create</code></h2>​Request Body<ul><li>​<code>name</code> — Required. The name of the event.</li><li>​<code>start_at</code> — The start of the event in a Date ISO string.</li><li>​<code>cover_url</code> — Valid URL for the event cover photo. Restricted to Unsplash or an image on the Luma CDN. </li><li>​<code>duration_minutes</code> — The length of your event in minutes.</li><li>​<code>require_rsvp</code> — Require guests to register in order to see meeting information.</li><li>​<code>require_rsvp_approval</code> — Require host approval for a guest to see meeting information. If true, <code>require_rsvp</code> must also be true.</li><li>​<code>zoom_creation_method</code> — <code>&quot;created-automatically&quot; | &quot;existing-attached&quot; | &quot;manually-entered&quot;</code>. If you want us to create a Zoom on your account, choose <code>&quot;created-automatically&quot;</code>.</li><li>​<code>zoom_meeting_id</code></li><li>​<code>zoom_meeting_password</code></li><li>​<code>zoom_meeting_url</code></li><li>​<code>zoom_session_type</code> — <code>&quot;meeting&quot; | &quot;webinar&quot;</code></li><li>​<code>community_api_id</code> — The community that the event should be linked to. </li></ul>​Response Object<ul><li>​<code>api_id</code> — The ID of the event.</li></ul><h2>​<code>GET /user</code></h2>​Returns the current user if authenticated. If not authenticated, returns a 401.<h2>​<code>POST /user/audience/import</code></h2>​Import members into your audience. Tag existing members.​Request Body<ul><li>​<code>emails</code> — Required. An array of emails.</li><li>​<code>member_tag_api_ids</code> — Optional. An array of tag IDs to apply to the members.</li></ul>​Response Object<ul><li>​This route returns an empty object on success.</li></ul></div>

Luma API (Beta)

While Substack is a great way to distribute newsletters, it doesn't provide a good tools for developing an active community. 
We've built  Luma Communities  to encourage member interaction and to foster relationships between members. 
With a Luma Community, your members will be able to access private content, find other members, meet in events, and even chat within Luma. 
Here is a playbook for transferring from Substack to Luma:

Luma Profile + Luma Community
While Substack gives you one page for public and private content, Luma gives you different pages for your public and private content. 
Your public content will live on your Luma Profile ( here is an example ) while your private content and community home will live in your Luma Community. 

Step 1 — Set Up Your Luma Profile
You can customize your Luma Profile here:  lu.ma/me
Be sure to add a Posts block to your Luma Profile since that's how we will display your public blog posts. 
Bonus — Import Old Substack Posts
For now, contact us about this! We will help you do this on the backend. 

Step 2 — Import Your Free Audience
You can export your entire Substack list from your Substack Dashboard → Subscribers tab. 

Once you have exported those members, you can import them into your Luma Audience here:  lu.ma/home/audience
When you import members, you can add the tag "Substack" so you know where they came from. 
Once you have imported members into your audience, you will be able to send them newsletters from within Luma. 

Step 3 — Set Up Your Luma Community
Your  Luma Community  gives your paying members a warm home where they can find private content, register for events, and meet other members. 
Create a community and populate it by adding events, adding to the library, and customizing member profiles. 
Set up payment options on the Settings → Payment tab so that new members can pay for access to your community. 
Add a link to your Luma Community on your Luma Profile so that people can discover your Luma Community. 

Step 4 — Import Paying Members to Your Community
We will help with this for now! Contact us 😀

Step 5 — Launch
Now you'll want to let your members know that you've launched on Luma

<div class="mirror-content" data-reactroot="">​While Substack is a great way to distribute newsletters, it doesn&#x27;t provide a good tools for developing an active community. ​We&#x27;ve built <a href="https://lu.ma/communities" target="_blank" rel="nofollow noopener">Luma Communities</a> to encourage member interaction and to foster relationships between members. ​With a Luma Community, your members will be able to access private content, find other members, meet in events, and even chat within Luma. ​Here is a playbook for transferring from Substack to Luma:​<h2>​Luma Profile + Luma Community</h2>​While Substack gives you one page for public and private content, Luma gives you different pages for your public and private content. ​Your public content will live on your Luma Profile (<a href="https://lu.ma/lolitataub" target="_blank" rel="nofollow noopener">here is an example</a>) while your private content and community home will live in your Luma Community. ​<h2>​Step 1 — Set Up Your Luma Profile</h2>​You can customize your Luma Profile here: <a href="http://lu.ma/me" target="_blank" rel="nofollow noopener">lu.ma/me</a>​Be sure to add a Posts block to your Luma Profile since that&#x27;s how we will display your public blog posts. ​Bonus — Import Old Substack Posts​For now, contact us about this! We will help you do this on the backend. ​<h2>​Step 2 — Import Your Free Audience</h2>​You can export your entire Substack list from your Substack Dashboard → Subscribers tab. <img alt="CleanShot 2021-08-02 at 13.52.18.gif" title="CleanShot 2021-08-02 at 13.52.18.gif" src="https://uploads.linear.app/9f5f1290-2248-4b9b-acd2-1c80cf83eb0f/50468395-18bd-4520-8ecf-7659f5ef8f07/773542a8-934c-488d-9fa1-7e5672a74b8d"/>​Once you have exported those members, you can import them into your Luma Audience here: <a href="http://lu.ma/home/audience" target="_blank" rel="nofollow noopener">lu.ma/home/audience</a>​When you import members, you can add the tag &quot;Substack&quot; so you know where they came from. ​Once you have imported members into your audience, you will be able to send them newsletters from within Luma. ​<h2>​Step 3 — Set Up Your Luma Community</h2>​Your <a href="https://lu.ma/communities" target="_blank" rel="nofollow noopener">Luma Community</a> gives your paying members a warm home where they can find private content, register for events, and meet other members. ​Create a community and populate it by adding events, adding to the library, and customizing member profiles. ​Set up payment options on the Settings → Payment tab so that new members can pay for access to your community. ​Add a link to your Luma Community on your Luma Profile so that people can discover your Luma Community. ​<h2>​Step 4 — Import Paying Members to Your Community</h2>​We will help with this for now! Contact us 😀​<h2>​Step 5 — Launch</h2>​Now you&#x27;ll want to let your members know that you&#x27;ve launched on Luma<blockquote>​​Hi, I wanted to share an exciting development for the community. ​I&#x27;ve been looking for a space where we can connect beyond this blog. And I think I&#x27;ve found the place with our new community on lu.ma! ​In our private lu.ma, you&#x27;ll be able to find private posts, browse the member directory, see upcoming events, and chat with other members. ​If you are a paying member of the Substack, you&#x27;ll get an invite to the Luma Community shortly or you can sign in with your email here (todo link this).​If you are a free member, you don&#x27;t need to do anything! You&#x27;ll still get emails as usual, but they&#x27;ll be coming from Luma. ​Really excited!</blockquote>​</div>

CleanShot 2021-08-02 at 13.52.18.gif

Playbook — Migrate from Substack to Luma

​ Luma is hiring software engineers + designers. If you refer someone who joins, we'll give you $15k — shoot us  an email . We've come really close to giving this bonus out. So will you be the first one??
OK so it looks like these formerly weekly updates are now turning monthly!
Luma Releases — Paid Communities + Notifications

Two weeks ago we launched  Paid Communities  on Luma. This lets community leaders charge for access to their online course, paid community, or conference.
This week (tomorrow) we are going to launch a Notifications tab inside the community and a beta of Community Cohorts which has some exciting beta testers including  First Round .
At the start of this quarter, we outlined a bi-weekly release schedule and it's helping to keep ourselves on track!
Paddington Movie Review
My housemates in New York have been telling me that  Paddington 2 is the best movie of all time . This is technically true according to the arcane rules that Rotten Tomatoes uses. But it convinced me enough to watch Paddington.
And I loved Paddington and Paddington 2!
The movie is ostensibly about a young bear whose jungle home gets damaged in an earthquake and decides to start a new life in London. But it's really about how you should try to see the best in people and how overwhelming kindness can surmount almost all barriers. 
Paddington is also a pretty good follow on Twitter with quips like " It has been a nice weekend. "
Running ➡️ 5K and Mile PRs!
I've been running more mileage, more consistently. And I've also run a couple races!
I ran two PRs:
Mile — 5:15 in Prospect Park, Brooklyn
5k — 19:12 in Rockaway Beach. We took the Ferry back which is basically a sea tour of New York and it only costed $2.50

After the 5k in Rockaway! (This was actually a really fast race and the winner came in at 14 minutes. I was just happy he didn't lap me...)

After the mile in Prospect Park with Ben who was really close to me in time but unfortunately had to run in a different heat.
In a couple weeks, I'm going to run the Brooklyn Mile and my stretch goal is a 5:05. 
Goals Coming Up
Each week, I've been setting goals for the upcoming week. But I haven't been emailing those out like in the past. 
Right now the goals are revolving around the following:
Running — Steadily increasing my weekly mileage which is now at 20 miles / week
Luma  Releases — build fun new stuff
Hiring — hiring engineers, designers, + a community lead

<div class="mirror-content" data-reactroot="">​​Luma is hiring software engineers + designers. If you refer someone who joins, we&#x27;ll give you $15k — shoot us <a href="http://mailto:jobs@lu.ma/" target="_blank" rel="nofollow noopener">an email</a>. We&#x27;ve come really close to giving this bonus out. So will you be the first one??​OK so it looks like these formerly weekly updates are now turning monthly!<h2>​Luma Releases — Paid Communities + Notifications</h2><img width="374" src="https://cdn.lu.ma/cdn-cgi/image/format=auto,dpr=2,quality=85,width=374/editor-images/xd/29bbd139-40cd-44f0-8064-10b35a738a2f"/>​Two weeks ago we launched <a href="https://lu.ma/p/A2fINHnhSWMH0DC/Paid-Communities" target="_blank" rel="nofollow noopener">Paid Communities</a> on Luma. This lets community leaders charge for access to their online course, paid community, or conference.​This week (tomorrow) we are going to launch a Notifications tab inside the community and a beta of Community Cohorts which has some exciting beta testers including <a href="https://firstround.com/" target="_blank" rel="nofollow noopener">First Round</a>.​At the start of this quarter, we outlined a bi-weekly release schedule and it&#x27;s helping to keep ourselves on track!<h2>​Paddington Movie Review</h2>​My housemates in New York have been telling me that <a href="https://www.vanityfair.com/hollywood/2021/04/paddington-2-just-became-the-best-reviewed-movie-of-all-time" target="_blank" rel="nofollow noopener">Paddington 2 is the best movie of all time</a>. This is technically true according to the arcane rules that Rotten Tomatoes uses. But it convinced me enough to watch Paddington.​And I loved Paddington and Paddington 2!​The movie is ostensibly about a young bear whose jungle home gets damaged in an earthquake and decides to start a new life in London. But it&#x27;s really about how you should try to see the best in people and how overwhelming kindness can surmount almost all barriers. ​Paddington is also a pretty good follow on Twitter with quips like &quot;<a href="https://twitter.com/paddingtonbear/status/1363543525791531008" target="_blank" rel="nofollow noopener">It has been a nice weekend.</a>&quot;<h2>​Running ➡️ 5K and Mile PRs!</h2>​I&#x27;ve been running more mileage, more consistently. And I&#x27;ve also run a couple races!​I ran two PRs:<ul><li>​Mile — 5:15 in Prospect Park, Brooklyn</li><li>​5k — 19:12 in Rockaway Beach. We took the Ferry back which is basically a sea tour of New York and it only costed $2.50</li></ul><img width="439" src="https://cdn.lu.ma/cdn-cgi/image/format=auto,dpr=2,quality=85,width=439/editor-images/4p/0086f006-789f-451f-a3d7-2f3444d70b62"/>​After the 5k in Rockaway! (This was actually a really fast race and the winner came in at 14 minutes. I was just happy he didn&#x27;t lap me...)<img width="300" src="https://cdn.lu.ma/cdn-cgi/image/format=auto,dpr=2,quality=85,width=300/editor-images/ck/0fe07bf3-262e-4bc4-9e08-795e23d1c9c3"/>​After the mile in Prospect Park with Ben who was really close to me in time but unfortunately had to run in a different heat.​In a couple weeks, I&#x27;m going to run the Brooklyn Mile and my stretch goal is a 5:05. <h2>​Goals Coming Up</h2>​Each week, I&#x27;ve been setting goals for the upcoming week. But I haven&#x27;t been emailing those out like in the past. ​Right now the goals are revolving around the following:<ul><li>​Running — Steadily increasing my weekly mileage which is now at 20 miles / week</li><li>​Luma <ul><li>​Releases — build fun new stuff</li><li>​Hiring — hiring engineers, designers, + a community lead</li></ul></li></ul>​</div>

VP#105 — Luma Releases, Paddington, Running

Luma is hiring software engineers + designers. If you refer someone who joins, we'll give you $15k — shoot us  an email .  
Luma's Tech Stack

I wrote a  blog post  about the tech we use at Luma
My Mom Visited NYC
My mom visited New York for about a week! We spent a lot of time walking around the city — Dumbo, West Village, High Line, Little Island + Soho.
And we had a really nice lunch at  Via Carota  (pictured below)

Running Goal ➡️ Sub 5 Min Mile
I'm once again focused on my goal to run a sub-5 minute mile. I did a benchmark run on Saturday and ran a 5:55 mile followed by a 6:05 mile with 6 minutes rest between the two. 
In the past, I've gotten too excited and ramped up my training too quickly and injured myself. So this time, I'm doing my best to gradually ramp up my mileage around 10-15% per week.
Goals for the Week
Run 16-17 miles — not more!
Launch new Luma Communities landing page
Improve recruiting pipeline (see above re $15k referral!)

<div class="mirror-content" data-reactroot="">​Luma is hiring software engineers + designers. If you refer someone who joins, we&#x27;ll give you $15k — shoot us <a href="http://mailto:jobs@lu.ma" target="_blank" rel="nofollow noopener">an email</a>. <h2>​Luma&#x27;s Tech Stack</h2><img src="https://cdn.lu.ma/cdn-cgi/image/format=auto,dpr=2,quality=85,width=800/editor-images/l5/5e20183b-6b03-4f79-afe9-bf37b2f056b1"/>​I wrote a <a href="https://lu.ma/p/zPamspcICt0f3q9/Lumas-Tech-Stack" target="_blank" rel="nofollow noopener">blog post</a> about the tech we use at Luma<h2>​My Mom Visited NYC</h2>​My mom visited New York for about a week! We spent a lot of time walking around the city — Dumbo, West Village, High Line, Little Island + Soho.​And we had a really nice lunch at <a href="https://www.viacarota.com/" target="_blank" rel="nofollow noopener">Via Carota</a> (pictured below)<img src="https://cdn.lu.ma/cdn-cgi/image/format=auto,dpr=2,quality=85,width=800/editor-images/60/901cbed1-3b1b-4ad5-bd7c-f55acad4502f"/><h2>​Running Goal ➡️ Sub 5 Min Mile</h2>​I&#x27;m once again focused on my goal to run a sub-5 minute mile. I did a benchmark run on Saturday and ran a 5:55 mile followed by a 6:05 mile with 6 minutes rest between the two. ​In the past, I&#x27;ve gotten too excited and ramped up my training too quickly and injured myself. So this time, I&#x27;m doing my best to gradually ramp up my mileage around 10-15% per week.<h2>​Goals for the Week</h2><ul><li>​Run 16-17 miles — not more!</li><li>​Launch new Luma Communities landing page</li><li>​Improve recruiting pipeline (see above re $15k referral!)</li></ul>​</div>

VP#104 — Tech, Mom, Running

Hi! We get a lot of questions about what we use for Luma’s tech stack. I’ll share the major technologies we chose along with some of the reasons why we chose them.
And if you want to work with our tech stack, we are hiring . 😍

Frontend
React 
 I started programming with React in 2015 and it felt like a revelation. It still does. I can’t imagine building with another view library today. 
Next.js 
 Next.js provides a great framework for code splitting, server rendering, and a ton of other small things. We’re really happy with where Next.js is today and the direction they are heading in. 
 
 Next.js also handles most of the Babel / Webpack config for us so that we haven’t had to spent time optimizing our bundle.
styled-jsx + SCSS 
 We started with CSS Modules but it was frustrating to have a different CSS file for every JS file. Whenever you wanted to move a component to a different file, you would need to figure out how to move the associated CSS file.
I looked at the various css-in-js options and chose styled-jsx because it seemed relatively well adopted and was made by the makers of Next.js. 
I am not confident in this decision. styled-jsx has a hard time styling nested components and emotion seems like it’s getting more popular. 
A Note on Design 
 We care tremendously about delivering a delightful experience. That means spending time iterating on the UX and actually working through edge cases. 
We use some Bootstrap utility classes but most of our CSS and design is done in house. In order to deliver a warm emotional response, we’ve built our own components and design system.

Backend
We chose Typescript for the frontend and backend for a few advantages:
Developers code in one language
We share code between the frontend / backend
We type our API routes
Koa web server 
 Express is the most popular Node.js web server but it hasn’t had a major release in 6 years which means it still has slightly awkward support for async / await . 
Koa was made by the maker of Express and is built with async / await as native primitives. Unfortunately Koa is not evolving or improving anymore. But it’s a minimalistic web server that gets the job done. 
I have been surprised by the lack of progress in the Node.js web server space. Today most of my hope rests on Deno which will rewrite the way that Node.js handles HTTP requests. 
Postgres DB 
 I love Postgres. While we briefly considered MySQL since my co-founder used it at Uber, we settled on Postgres since it’s the gold standard in DBs. 
We don’t use too many advanced Postgres features but we like JSONB support and have created a function to generate prefixed IDs .
Knex.js + Raw SQL 
 I’ve played with a lot of different JS ORMs. But none of them are at the maturity of ActiveRecord or SQLAlchemy. Once your project gets to a certain size the JS ORM starts hurting you more than helping you. 
We use Knex.js as a query builder but we often resort to just writing Raw SQL. This has made our API routes obviously performant and has not slowed down development. 
I think there is promise in Prisma but at this point it feels too magical and heavy for us to adopt it.

Service Architecture
We try to avoid having too many services unless the services have very different resource needs or performance characteristics. 
Here are the different perf characteristics of our services:
Web — serve HTML / JS / CSS, serve it fast
API — serve JSON in < 1s, longer responses should be an async job
Async Job System — we use a queuing + worker system to process jobs asynchronously
Screenshots — we use Puppeteer to generate screenshots, this can crash and take longer than a second so we don’t group it with the API
Realtime — we use Centrifugo as a service to manage websockets and send realtime updates to clients

Prod Infrastructure
We launched Luma on Vercel and Render . While I love both of those services and they allowed us to move insanely quickly, we ended up switching over our production infrastructure entirely to AWS. 
We were attracted to AWS’s reliability and the ability to have all of our services in one place. Here is how we use AWS:
ECS + Fargate 
 Our services all use ECS to spin up / down containers. There was a steep learning curve when I had to understand Services, Tasks, Task Definitions, IAM, Security Group and a million other terms. But now it seems pretty easy. 
Aurora for the DB 
 Aurora is amazing. It separates compute and storage so that we haven’t had to worry about scaling our database up. It also gives us peace of mind with point in time backups. 
Github Actions for Deploys and Tests 
 We run a test suite with every Pull Request that must pass before the PR is merged into master.
For every commit on master, a Github Action updates our services. This means we might have 10+ deploys to production every day. Each deploy updates our services in a waterfall — DB is migrate, then the new API rolls out, then the web service. 
Vercel for Preview Deploys 
 Vercel deploys each pull request to a preview branch. This makes it really easy to test frontend changes on production. 
Sentry for Errors 
 Sentry feels like a React-level decision. Widely accepted and generally reliable.
Honeycomb for Instrumentation 
 We have set up a Koa middleware to wrap all of our API routes. This lets us see how long each route is taking or if we are seeing a spike in traffic.

<div class="mirror-content" data-reactroot="">​Hi! We get a lot of questions about what we use for Luma’s tech stack. I’ll share the major technologies we chose along with some of the reasons why we chose them.​And if you want to work with our tech stack, <a href="https://lu.ma/jobs" target="_blank" rel="nofollow noopener">we are hiring</a>. 😍​<h2>​Frontend</h2>​<a href="https://reactjs.org/" target="_blank" rel="nofollow noopener">React</a> I started programming with React in 2015 and it felt like a revelation. It still does. I can’t imagine building with another view library today. ​<a href="https://nextjs.org/" target="_blank" rel="nofollow noopener">Next.js</a> Next.js provides a great framework for code splitting, server rendering, and a ton of other small things. We’re really happy with where Next.js is today and the direction they are heading in. Next.js also handles most of the Babel / Webpack config for us so that we haven’t had to spent time optimizing our bundle.​<a href="https://github.com/vercel/styled-jsx" target="_blank" rel="nofollow noopener">styled-jsx</a> + SCSS We started with <a href="https://nextjs.org/docs/advanced-features/customizing-postcss-config#css-modules" target="_blank" rel="nofollow noopener">CSS Modules</a> but it was frustrating to have a different CSS file for every JS file. Whenever you wanted to move a component to a different file, you would need to figure out how to move the associated CSS file.​I looked at the various css-in-js options and chose styled-jsx because it seemed relatively well adopted and was made by the makers of Next.js. ​I am not confident in this decision. styled-jsx has a hard time styling nested components and <a href="https://emotion.sh/docs/introduction" target="_blank" rel="nofollow noopener">emotion</a> seems like it’s getting more popular. ​A Note on Design We care tremendously about delivering a delightful experience. That means spending time iterating on the UX and actually working through edge cases. ​We use some Bootstrap utility classes but most of our CSS and design is done in house. In order to deliver a warm emotional response, we’ve built our own components and design system.​<h2>​Backend</h2>​We chose Typescript for the frontend and backend for a few advantages:<ul><li>​Developers code in one language</li><li>​We share code between the frontend / backend</li><li>​We <a href="https://lu.ma/p/wPYEf0aW0uok4PS/We-made-our-development-10x-faster-by-typing-our-API-interface" target="_blank" rel="nofollow noopener">type our API routes</a></li></ul>​<a href="https://koajs.com/" target="_blank" rel="nofollow noopener">Koa</a> web server Express is the most popular Node.js web server but it hasn’t had a major release in 6 years which means it still has slightly awkward support for <code>async</code> / <code>await</code>. ​Koa was made by the maker of Express and is built with <code>async</code> / <code>await</code> as native primitives. Unfortunately Koa is not evolving or improving anymore. But it’s a minimalistic web server that gets the job done. ​I have been surprised by the lack of progress in the Node.js web server space. Today most of my hope rests on <a href="https://deno.land/" target="_blank" rel="nofollow noopener">Deno</a> which will rewrite the way that Node.js handles HTTP requests. ​<a href="https://www.postgresql.org/" target="_blank" rel="nofollow noopener">Postgres</a> DB I love Postgres. While we briefly considered MySQL since my <a href="https://lu.ma/d" target="_blank" rel="nofollow noopener">co-founder</a> used it at Uber, we settled on Postgres since it’s the gold standard in DBs. ​We don’t use too many advanced Postgres features but we like JSONB support and have created a function to <a href="https://stackoverflow.com/a/41988979/1411726" target="_blank" rel="nofollow noopener">generate prefixed IDs</a>.​<a href="http://knexjs.org/" target="_blank" rel="nofollow noopener">Knex.js</a> + Raw SQL I’ve played with a lot of different JS ORMs. But none of them are at the maturity of ActiveRecord or SQLAlchemy. Once your project gets to a certain size the JS ORM starts hurting you more than helping you. ​We use Knex.js as a query builder but we often resort to just writing Raw SQL. This has made our API routes obviously performant and has not slowed down development. ​I think there is promise in Prisma but at this point it feels too magical and heavy for us to adopt it.​<h2>​Service Architecture</h2>​We try to avoid having too many services unless the services have very different resource needs or performance characteristics. ​Here are the different perf characteristics of our services:<ul><li>​Web — serve HTML / JS / CSS, serve it fast</li><li>​API — serve JSON in &lt; 1s, longer responses should be an async job</li><li>​Async Job System — we use a queuing + worker system to process jobs asynchronously</li><li>​Screenshots — we use Puppeteer to generate screenshots, this can crash and take longer than a second so we don’t group it with the API</li><li>​Realtime — we use Centrifugo as a service to manage websockets and send realtime updates to clients</li></ul>​<h2>​Prod Infrastructure</h2>​We launched Luma on <a href="https://vercel.com/luma-team" target="_blank" rel="nofollow noopener">Vercel</a> and <a href="https://render.com/" target="_blank" rel="nofollow noopener">Render</a>. While I love both of those services and they allowed us to move insanely quickly, we ended up switching over our production infrastructure entirely to AWS. ​We were attracted to AWS’s reliability and the ability to have all of our services in one place. Here is how we use AWS:​ECS + Fargate Our services all use ECS to spin up / down containers. There was a steep learning curve when I had to understand Services, Tasks, Task Definitions, IAM, Security Group and a million other terms. But now it seems pretty easy. ​Aurora for the DB Aurora is amazing. It separates compute and storage so that we haven’t had to worry about scaling our database up. It also gives us peace of mind with point in time backups. ​Github Actions for Deploys and Tests We run a test suite with every Pull Request that must pass before the PR is merged into master.​For every commit on master, a Github Action updates our services. This means we might have 10+ deploys to production every day. Each deploy updates our services in a waterfall — DB is migrate, then the new API rolls out, then the web service. ​<a href="https://vercel.com/" target="_blank" rel="nofollow noopener">Vercel</a> for Preview Deploys Vercel deploys each pull request to a preview branch. This makes it really easy to test frontend changes on production. ​<a href="https://sentry.io/" target="_blank" rel="nofollow noopener">Sentry</a> for Errors Sentry feels like a React-level decision. Widely accepted and generally reliable.​<a href="https://www.honeycomb.io/" target="_blank" rel="nofollow noopener">Honeycomb</a> for Instrumentation We have set up a Koa middleware to wrap all of our API routes. This lets us see how long each route is taking or if we are seeing a spike in traffic. </div>

Luma's Tech Stack

Build Club 1 — Making Games in the Browser
I’m excited to welcome Nikhilesh Sigatapu as the first guest of  Build Club . He’s one of the most interesting builders I know and is always pushing the boundaries of what’s possible.

 Join us next Thursday at 4pm PT / 7pm ET for a 30-40 minute presentation and hanging out afterwards. 
Register for Build Club

Life in New York

Life has been good in New York! Last night I realized I hadn’t eaten dinner so I went to a nearby deli and grabbed a sandwich… at 1am. 
I’m also getting tons of opportunities to meet people in person which is really refreshing. The above photo was taken by the kind  Eugene Wei  who spent time with me talking about the next internet over tapas and beers.
Luma Community Preview
At Luma, we’ve helped creators run events, send newsletters, and build a personal websites. As part of that evolution, we’re now building a space where you can bring people together. 
We’ve been talking to Eugene a lot about “self-describing UIs” so instead of explaining what a Luma community is — I’ll share a screenshot of our internal community.

The hosts in our beta program are using it to teach online classes, run alumni groups, and host professional networks. 
If you’re interested in learning more or getting into the beta, reply to this email or join our  beta tester community . 😎
Elephant in the Brain — Review
Over the past month, I’ve read  The Elephant in the Brain  by Robin Hanson and Kevin Simler. The book argues that we have two types of motives — stated motives and hidden motives.
Take an example like giving to charity. The stated motive is that we want to help people and we are a good person. But the hidden motive is that we want to appear kind to others who are considering choosing us as a friend or ally. 
The book explores how this type of signaling explains many different types of human behavior. While social signaling isn’t a new idea, the book makes compelling arguments that give explanations for artists, religious martyrs, and press spokespeople. 
It’s worth a read if you’d like to have a more mechanistic understanding of human psychology.

<div class="mirror-content" data-reactroot=""><h2>​Build Club 1 — Making Games in the Browser</h2>​I’m excited to welcome Nikhilesh Sigatapu as the first guest of <a href="https://lu.ma/p/RyFpGlpw82ppLdG/VP#102-I-m-Moving-to-New-York-+-Announcing-Build-Club" target="_blank" rel="nofollow noopener">Build Club</a>. He’s one of the most interesting builders I know and is always pushing the boundaries of what’s possible.<img src="https://cdn.lu.ma/editor-images/is/c184c747-cba2-45e4-ac12-e405eec271a1"/>​ Join us next Thursday at 4pm PT / 7pm ET for a 30-40 minute presentation and hanging out afterwards. <div style="text-align:center;margin-bottom:1rem"><a href="https://lu.ma/build-club-1" class="btn" style="background-color:#f68291;color:#ffffff;border:1px solid transparent;padding:0.375rem 0.75rem;border-radius:4px;font-size:1rem !important;line-height:1.5 !important;word-break:break-word;display:inline-block;text-align:center;font-weight:bold;text-decoration:none !important" target="_blank" rel="nofollow noopener">Register for Build Club</a></div>​<h2>​Life in New York</h2><img src="https://cdn.lu.ma/editor-images/1o/0f7f187e-93b4-477e-a28c-7695ea136279"/>​Life has been good in New York! Last night I realized I hadn’t eaten dinner so I went to a nearby deli and grabbed a sandwich… at 1am. ​I’m also getting tons of opportunities to meet people in person which is really refreshing. The above photo was taken by the kind <a href="https://www.eugenewei.com/" target="_blank" rel="nofollow noopener">Eugene Wei</a> who spent time with me talking about the next internet over tapas and beers.<h2>​Luma Community Preview</h2>​At Luma, we’ve helped creators run events, send newsletters, and build a personal websites. As part of that evolution, we’re now building a space where you can bring people together. ​We’ve been talking to Eugene a lot about “self-describing UIs” so instead of explaining what a Luma community is — I’ll share a screenshot of our internal community.<img src="https://cdn.lu.ma/editor-images/3x/9a58fe2b-4be1-41ab-a222-d7a7de886f5f"/>​The hosts in our beta program are using it to teach online classes, run alumni groups, and host professional networks. ​If you’re interested in learning more or getting into the beta, reply to this email or join our <a href="https://lu.ma/community/com-iacPIKhTqGgBDwg" target="_blank" rel="nofollow noopener">beta tester community</a>. 😎<h2>​Elephant in the Brain — Review</h2>​Over the past month, I’ve read <a href="https://www.amazon.com/Elephant-Brain-Hidden-Motives-Everyday/dp/0190495995" target="_blank" rel="nofollow noopener">The Elephant in the Brain</a> by Robin Hanson and Kevin Simler. The book argues that we have two types of motives — stated motives and hidden motives.​Take an example like giving to charity. The stated motive is that we want to help people and we are a good person. But the hidden motive is that we want to appear kind to others who are considering choosing us as a friend or ally. ​The book explores how this type of signaling explains many different types of human behavior. While social signaling isn’t a new idea, the book makes compelling arguments that give explanations for artists, religious martyrs, and press spokespeople. ​It’s worth a read if you’d like to have a more mechanistic understanding of human psychology. </div>

VP#103 — First Session of Build Club

VP#102 — Moving to New York + Announcing Build Club
Quick aside: we will give you $15k if you successfully refer a software engineer to Luma. (This also counts if you refer yourself!) Send intros to  jobs@lu.ma

I’m Moving to New York!
I’m writing this right now while on a plane from SFO to JFK. Last night I packed up everything I own into two suitcases and a backpack. 

What I hope to find in New York:
An endless stream of new faces and places
Walkable neighborhoods 
Ambitious (non tech) people 
Late night food 
Energy
Perspective 
What I hope to do in New York:
Find an airy, light filled office where we build Luma and host events
Join or create basketball and running groups
Host bi-weekly dinners / meetups around some central theme 
Get involved with pro housing and pro transit groups 
I’ve only spent about a month cumulatively in New York. So I’m excited and apprehensive to chart new territory. 
If you know of anyone I should meet in NYC, let me know! 

Announcing Build Club
I have so many friends that are building really cool things from open source projects to LED art projects to crypto projects to billion dollar companies. (You are probably one of them!) 
I’ve learned so much from these people and I am constantly asking them for help. But most of these people haven’t spoken at a single conference and you wouldn’t know how awesome and helpful they are.
So I’m starting this club to bring people together, host events, and see what we can build together. 
I’m planning on starting with a series of short tech talks online. We have a bunch of great hosts already confirmed. 
If you are interested in speaking or participating or just have thoughts:
Fill Out This Form

And if you have an idea for a better name that Build Club, let me know…

<div class="mirror-content" data-reactroot=""><h1>​VP#102 — Moving to New York + Announcing Build Club</h1>​Quick aside: we will give you $15k if you successfully refer a software engineer to Luma. (This also counts if you refer yourself!) Send intros to <a href="https://mailto:jobs@lu.ma" target="_blank" rel="nofollow noopener">jobs@lu.ma</a>​<h2>​I’m Moving to New York!</h2>​I’m writing this right now while on a plane from SFO to JFK. Last night I packed up everything I own into two suitcases and a backpack. <img src="https://cdn.lu.ma/editor-images/bj/984c9b64-d62d-4739-b737-8ae0eb5d1866"/>​What I hope to find in New York:<ul><li>​An endless stream of new faces and places</li><li>​Walkable neighborhoods </li><li>​Ambitious (non tech) people </li><li>​Late night food </li><li>​Energy</li><li>​Perspective </li></ul>​What I hope to do in New York:<ul><li>​Find an airy, light filled office where we build Luma and host events</li><li>​Join or create basketball and running groups</li><li>​Host bi-weekly dinners / meetups around some central theme </li><li>​Get involved with pro housing and pro transit groups </li></ul>​I’ve only spent about a month cumulatively in New York. So I’m excited and apprehensive to chart new territory. ​If you know of anyone I should meet in NYC, let me know! ​<h2>​Announcing Build Club</h2>​I have so many friends that are building really cool things from open source projects to LED art projects to crypto projects to billion dollar companies. (You are probably one of them!) ​I’ve learned so much from these people and I am constantly asking them for help. But most of these people haven’t spoken at a single conference and you wouldn’t know how awesome and helpful they are.​So I’m starting this club to bring people together, host events, and see what we can build together. ​I’m planning on starting with a series of short tech talks online. We have a bunch of great hosts already confirmed. ​If you are interested in speaking or participating or just have thoughts:<div style="text-align:center;margin-bottom:1rem"><a href="https://airtable.com/shrG0RidueiHjDITf" class="btn" style="background-color:#f68291;color:#ffffff;border:1px solid transparent;padding:0.375rem 0.75rem;border-radius:4px;font-size:1rem !important;line-height:1.5 !important;word-break:break-word;display:inline-block;text-align:center;font-weight:bold;text-decoration:none !important" target="_blank" rel="nofollow noopener">Fill Out This Form</a></div>​​And if you have an idea for a better name that Build Club, let me know…</div>

VP#102 — I'm Moving to New York + Announcing Build Club

Hi! For the last week I’ve been in Los Angeles at a house with 20 or so other startup people called  Launch House . We’re living in a big house nestled in the hills above West Hollywood.
It’s nearly the opposite of what I’ve been doing for the past year — holing up with a small group of friends to avoid COVID. At first, I thought it would take time to readjust to meeting people in person and going out. But in just a week, I can’t imagine going back to a world of masks and isolation.

Here’s the view from our balcony. It’s an amazing house in a location that reminds me why I wouldn’t want to live in LA (you have to drive everywhere). 
But LA has one thing that no other city has… the Lakers! I went to the Lakers game tonight and saw Anthony Davis and the Lakers thoroughly reck the Suns. 

It was really fun to see the Lakers live but the COVID rules are so strange. You need to prove vaccination or COVID test and still the stadium was at 5% capacity, enforced distancing, and required mask wearing. California really doesn’t want things to go back to normal. 
I’ll be in LA for the next week. And then I’m headed to New York. Exciting times ahead. 
Oh and by the way, we’re expanding Luma into a new community platform that pulls together your events, member directory, and more. Here’s a  sneak preview , let me know what you think.

<div class="mirror-content" data-reactroot="">​Hi! For the last week I’ve been in Los Angeles at a house with 20 or so other startup people called <a href="https://www.launchhouse.co/" target="_blank" rel="nofollow noopener">Launch House</a>. We’re living in a big house nestled in the hills above West Hollywood.​It’s nearly the opposite of what I’ve been doing for the past year — holing up with a small group of friends to avoid COVID. At first, I thought it would take time to readjust to meeting people in person and going out. But in just a week, I can’t imagine going back to a world of masks and isolation.<img src="https://cdn.lu.ma/editor-images/eq/f8b86f82-47b4-4e17-86f6-0d97342f7bcf"/>​Here’s the view from our balcony. It’s an amazing house in a location that reminds me why I wouldn’t want to live in LA (you have to drive everywhere). ​But LA has one thing that no other city has… the Lakers! I went to the Lakers game tonight and saw Anthony Davis and the Lakers thoroughly reck the Suns. <img src="https://cdn.lu.ma/editor-images/6s/0ec44af8-139e-4177-b96b-bfd75f437b80"/>​It was really fun to see the Lakers live but the COVID rules are so strange. You need to prove vaccination or COVID test and still the stadium was at 5% capacity, enforced distancing, and required mask wearing. California really doesn’t want things to go back to normal. ​I’ll be in LA for the next week. And then I’m headed to New York. Exciting times ahead. ​Oh and by the way, we’re expanding Luma into a new community platform that pulls together your events, member directory, and more. Here’s a <a href="https://www.loom.com/share/a10b218fd4b7407080ce2f42c5697edb" target="_blank" rel="nofollow noopener">sneak preview</a>, let me know what you think.</div>

VP#101 — LA, Startups, and Anthony Davis

For now, I'm going to call this weekly update VP#issue-num. I don't know how many I’ve sent so far, so let's just call this 100.

I first moved to San Francisco in January 2013 to intern at a company called Cardspring.
Cardspring forgot to book intern housing so I ended up in a dorm-style apartment in the Castro that violated nearly every occupancy law. The apartment was crappy and the people were weird, but that didn't matter. San Francisco felt alive. 
Everyone I met had a new idea, a new company, a new office, a new job. To live in San Francisco meant you were constantly changing and improving. Today, you are software engineer at a big company. Tomorrow, you work at a startup. Next year, you start your own company. 
Since my internship at Cardspring, I've lived in San Francisco for most of the last 7 years. And for the past 5 years I've lived with a group of incredibly kind, talented friends who have become closer than family.
But last night, I said goodbye to San Francisco. Tonight, I'm writing this letter from LA where I'll be for the next two weeks. In two weeks on May 17th, I'll be moving to New York City. 
There are a few reasons why I left San Francisco that I’d be happy to get into if you are interested. But I wasn’t pushed out of San Francisco, I was pulled to New York. By the allure of living with 10 million other people and living in walkable neighborhoods.
I’ve always wanted to live in New York and now it's finally possible. I’m both excited and nervous about the future. It’s a good feeling.
So if you're in LA for the next two weeks or in New York any time after that, let me know. I would love to see you.

<div class="mirror-content" data-reactroot="">​For now, I&#x27;m going to call this weekly update VP#issue-num. I don&#x27;t know how many I’ve sent so far, so let&#x27;s just call this 100.<img src="https://cdn.lu.ma/editor-images/h5/dc402b2a-ce17-4522-8195-473c1e59e846"/>​I first moved to San Francisco in January 2013 to intern at a company called Cardspring.​Cardspring forgot to book intern housing so I ended up in a dorm-style apartment in the Castro that violated nearly every occupancy law. The apartment was crappy and the people were weird, but that didn&#x27;t matter. San Francisco felt alive. ​Everyone I met had a new idea, a new company, a new office, a new job. To live in San Francisco meant you were constantly changing and improving. Today, you are software engineer at a big company. Tomorrow, you work at a startup. Next year, you start your own company. ​Since my internship at Cardspring, I&#x27;ve lived in San Francisco for most of the last 7 years. And for the past 5 years I&#x27;ve lived with a group of incredibly kind, talented friends who have become closer than family.​But last night, I said goodbye to San Francisco. Tonight, I&#x27;m writing this letter from LA where I&#x27;ll be for the next two weeks. In two weeks on May 17th, I&#x27;ll be moving to New York City. ​There are a few reasons why I left San Francisco that I’d be happy to get into if you are interested. But I wasn’t pushed out of San Francisco, I was pulled to New York. By the allure of living with 10 million other people and living in walkable neighborhoods.​I’ve always wanted to live in New York and now it&#x27;s finally possible. I’m both excited and nervous about the future. It’s a good feeling.​So if you&#x27;re in LA for the next two weeks or in New York any time after that, let me know. I would love to see you.</div>

VP#100 — Goodbye San Francisco

In previous, I’ve used the subject “Weekly Update YYYY-MM-DD — Title” which is a bit of a mouthful. I’m going to experiment with some new subjects. 
Ideally, I’ll come up with a cool name for this series…

Not Built for Paradise

I got back to SF on Saturday after spending a week in Maui. We went paddle boarding in the mornings and gathered for delicious, merry dinners in the evenings.
But I realized on that trip that I’m not built for paradise. While I enjoyed spending time in the sun and drinking pina coladas every afternoon, my body rejected it. 
Even though I put on sunscreen every morning and afternoon, my skin started to react to the sun at the end of the week. I think it was my body’s way of telling me that while spending hours in the sun is fun, a week is enough. 
So it’s time to get back to work.
I’ll be in San Francisco this week and then I’m headed to LA the week after. That’ll be the first week in a year that I’ll spend away from my COVID family. The next few months will have some big changes that I’m still processing…

Learning — Reflexivity and Ergodicity
I’ve spent today learning about  Reflexivity  and  Ergodicity .
Reflexivity is the concept that for certain systems, our beliefs can be self-fulfilling. If we believe a stock will go up, then it will go up. 
Ergodicity is a little more confusing and I’m still trying to grasp it. It’s the concept that the average over many different instances may not be the same as the average over time. …or something like that. I’m still trying to understand and hopefully I’ll have a more succinct explanation next time.

<div class="mirror-content" data-reactroot="">​In previous, I’ve used the subject “Weekly Update YYYY-MM-DD — Title” which is a bit of a mouthful. I’m going to experiment with some new subjects. ​Ideally, I’ll come up with a cool name for this series…​<h2>​Not Built for Paradise</h2><img src="https://cdn.lu.ma/editor-images/if/ac9464eb-d83c-4e63-896b-47211a21c0a7"/>​I got back to SF on Saturday after spending a week in Maui. We went paddle boarding in the mornings and gathered for delicious, merry dinners in the evenings.​But I realized on that trip that I’m not built for paradise. While I enjoyed spending time in the sun and drinking pina coladas every afternoon, my body rejected it. ​Even though I put on sunscreen every morning and afternoon, my skin started to react to the sun at the end of the week. I think it was my body’s way of telling me that while spending hours in the sun is fun, a week is enough. ​So it’s time to get back to work.​I’ll be in San Francisco this week and then I’m headed to LA the week after. That’ll be the first week in a year that I’ll spend away from my COVID family. The next few months will have some big changes that I’m still processing…​<h2>​Learning — Reflexivity and Ergodicity</h2>​I’ve spent today learning about <a href="https://twitter.com/VictorPontis/status/1386524308218859520" target="_blank" rel="nofollow noopener">Reflexivity</a> and <a href="https://twitter.com/VictorPontis/status/1386523892152365057" target="_blank" rel="nofollow noopener">Ergodicity</a>.​Reflexivity is the concept that for certain systems, our beliefs can be self-fulfilling. If we believe a stock will go up, then it will go up. ​Ergodicity is a little more confusing and I’m still trying to grasp it. It’s the concept that the average over many different instances may not be the same as the average over time. …or something like that. I’m still trying to understand and hopefully I’ll have a more succinct explanation next time. </div>

VP 21-04-25 — Not Built for Paradise

A Week in Hawaii
Hi, I’m writing to you with this view:

I’m in Maui for a week to celebrate the acquisition of my friend’s company. And if you walk to the beach in the photo, you can see the island of Lanai which is almost totally owned by Larry Ellison. 
So we are celebrating a friend’s success in tech while looking out and seeing an island bought by a database mogul. I’m sure there’s some metaphor or lesson here, but I can’t retrieve it. 
I’ll be in Maui for the rest of this week. I’m looking forward to paddle boarding, swimming, and drinking sugary cocktails. 
I would like to be running more here but I overextended myself with the marathon last week, so my knees are quite sore! I hope this can serve as a lesson that I really need to pace myself better. 

Achieving Your Goals
It’s hard to achieve the goals we set for ourselves. Most New Years resolutions fail. Gyms make most of their money off people who never go to the gym. The best universities have put their course content online for free but the completion rate of MOOCs is lower than 7%. 
We all have goals and skills we want to acquire, but again and again we struggle to achieve them. For the past 5 years, I’ve wanted to learn to draw and to dunk a basketball. 
While working on Luma, we’ve seen people come together to achieve their goals. To get a new job, workout together, learn how to cook. And it seems to be working. 
I’m thinking a lot more about how we achieve our goals. And what we can build in Luma to make it easier to achieve our goals. 
So my questions for you:
What skill do you want to improve or learn right now? Are you actively pursuing it?
Do you have any special tricks or habits that you use to achieve your goals?

<div class="mirror-content" data-reactroot=""><h2>​A Week in Hawaii</h2>​Hi, I’m writing to you with this view:<img src="https://cdn.lu.ma/editor-images/z2/a6908326-e2c5-41f3-a683-d3a6607a3286"/>​I’m in Maui for a week to celebrate the acquisition of my friend’s company. And if you walk to the beach in the photo, you can see the island of Lanai which is almost totally owned by Larry Ellison. ​So we are celebrating a friend’s success in tech while looking out and seeing an island bought by a database mogul. I’m sure there’s some metaphor or lesson here, but I can’t retrieve it. ​I’ll be in Maui for the rest of this week. I’m looking forward to paddle boarding, swimming, and drinking sugary cocktails. ​I would like to be running more here but I overextended myself with the marathon last week, so my knees are quite sore! I hope this can serve as a lesson that I really need to pace myself better. ​<h2>​Achieving Your Goals</h2>​It’s hard to achieve the goals we set for ourselves. Most New Years resolutions fail. Gyms make most of their money off people who never go to the gym. The best universities have put their course content online for free but the completion rate of MOOCs is lower than 7%. ​We all have goals and skills we want to acquire, but again and again we struggle to achieve them. For the past 5 years, I’ve wanted to learn to draw and to dunk a basketball. ​While working on Luma, we’ve seen people come together to achieve their goals. To get a new job, workout together, learn how to cook. And it seems to be working. ​I’m thinking a lot more about how we achieve our goals. And what we can build in Luma to make it easier to achieve our goals. ​So my questions for you:<ol><li>​What skill do you want to improve or learn right now? Are you actively pursuing it?</li><li>​Do you have any special tricks or habits that you use to achieve your goals? </li></ol></div>

Weekly Update 2021-04-18 — A Dispatch from Hawaii

First Marathon + Second Pfizer
I’ve spent most of the day today shuffling between my bed and the couch. I’m totally exhausted and it hurts to walk. That’s because I ran my first marathon yesterday!
We started at a friend’s house in Mill Valley, climbed over a mountain, through Muir Woods, down into Stinson Beach, and then looped back home.
I have been running a bit but I was definitely not prepared for this run. To put it in perspective, my weekly mileage for each of the past 10 weeks was less than a marathon…
But I had a great group of ~ 10 friends to run with and we stopped to stretch and eat every hour. And after mile 10 or so I stopped being able to feel my feet which means they stopped hurting!
Here was our route:


The day before the marathon I went with Dan (my co-founder) and Danqing (his girlfriend) to get our second Pfizer shot:

I was a bit nervous that I was going to get side effects from the shot because the marathon was starting 24 hours after I got the shot. But it all turned out alright! My arm was a little sore but otherwise I had no side effects. 
Luma Retreat
Earlier in the week (I guess this update is going in reverse chronological order), Dan and I went to Sonoma for a Luma planning retreat. 
As I wrote about in  my previous update , we started Luma ~ a year ago. And we both have a tendency to get focused on product, engineering, and customer relationships. So the retreat was a good time to lift up our heads and plan for the next year. 
The main thing that we came away with the retreat with is a new focus. We’ve built a lot of exciting things in the past year but they haven’t been tied together under one coherent vision. I’ll be writing more about this new vision this week.
Here’s a photo from the retreat so you can see just how serious and intense it was:

<div class="mirror-content" data-reactroot=""><h2>​First Marathon + Second Pfizer</h2>​I’ve spent most of the day today shuffling between my bed and the couch. I’m totally exhausted and it hurts to walk. That’s because I ran my first marathon yesterday!​We started at a friend’s house in Mill Valley, climbed over a mountain, through Muir Woods, down into Stinson Beach, and then looped back home.​I have been running a bit but I was definitely not prepared for this run. To put it in perspective, my weekly mileage for each of the past 10 weeks was less than a marathon…​But I had a great group of ~ 10 friends to run with and we stopped to stretch and eat every hour. And after mile 10 or so I stopped being able to feel my feet which means they stopped hurting!​Here was our route:<img src="https://cdn.lu.ma/editor-images/r2/e0fdb5fa-47db-409d-a08b-8936eb37202e"/>​​The day before the marathon I went with Dan (my co-founder) and Danqing (his girlfriend) to get our second Pfizer shot:<img src="https://cdn.lu.ma/editor-images/ff/bfb99f87-5683-4f1a-b7a4-05d9a8d5e2b9"/>​I was a bit nervous that I was going to get side effects from the shot because the marathon was starting 24 hours after I got the shot. But it all turned out alright! My arm was a little sore but otherwise I had no side effects. <h2>​Luma Retreat</h2>​Earlier in the week (I guess this update is going in reverse chronological order), Dan and I went to Sonoma for a Luma planning retreat. ​As I wrote about in <a href="https://lu.ma/p/lMgtd6UcK5XCDR4/Weekly-Update-2021-04-04" target="_blank" rel="nofollow noopener">my previous update</a>, we started Luma ~ a year ago. And we both have a tendency to get focused on product, engineering, and customer relationships. So the retreat was a good time to lift up our heads and plan for the next year. ​The main thing that we came away with the retreat with is a new focus. We’ve built a lot of exciting things in the past year but they haven’t been tied together under one coherent vision. I’ll be writing more about this new vision this week.​Here’s a photo from the retreat so you can see just how serious and intense it was:<img src="https://cdn.lu.ma/editor-images/7u/810cee68-6b33-4853-8c07-00e3d2755260"/>​</div>

Weekly Update 2021-04-11 — Luma Retreat, Pfizer #2, Marathon

Howdy! Luma Turns 1A year ago yesterday, we launched Luma (then ZmURL) on Techcrunch. I wrote a Twitter thread going over some of the high and low lights of the last year.With every anniversary or birthday, we’re spending more time reflecting on the past and planning for the future.We’ve seen 60,000 events on Luma in the last year. Each event brings people together in order to meet each other, participate an activity, or just spend time together. There’s a fundamental desire to connect with other people and the big social media platforms don’t do a good job of creating and nurturing new relationships.In the next year, we’re building Luma to help communities foster more meaningful relationships. That means more improvements to events, a member directory, spaces for people to meet each other, and much more.Back to BasketballThe hoops are back!San Francisco, against any reasonable scientific guidelines, has taken down the basketball hoops for most of COVID. But they are back now!I shot around for half an hour today and my shot is still kinda ok! Hopefully I’ll start to play some pickup games in the next couple weeks.Coming UpThis week I have a few exciting things:my 2nd vaccine shot (with the Danqings)a marathon in Marin on the day after my shot (hopefully I’m in shape to run it)a short Luma retreat (to spend more time on the planning I mentioned above)In terms of work, we’re planning on launching a member directory feature for Luma — a place where you can see bios and information about other members of your community. If you know of an online community that might benefit from a member directory, please let me know!

<div class="mirror-content" data-reactroot="">​Howdy! ​<h2>​Luma Turns 1</h2>​A year ago yesterday, we launched Luma (then ZmURL) on Techcrunch. I wrote a <a href="https://twitter.com/VictorPontis/status/1378869709773279233" target="_blank" rel="nofollow noopener">Twitter thread</a> going over some of the high and low lights of the last year.​With every anniversary or birthday, we’re spending more time reflecting on the past and planning for the future.​We’ve seen 60,000 events on Luma in the last year. Each event brings people together in order to meet each other, participate an activity, or just spend time together. There’s a fundamental desire to connect with other people and the big social media platforms don’t do a good job of creating and nurturing new relationships.​In the next year, we’re building Luma to help communities foster more meaningful relationships. That means more improvements to events, a member directory, spaces for people to meet each other, and much more.​<h2>​Back to Basketball</h2>​The hoops are back!​San Francisco, against any reasonable scientific guidelines, has taken down the basketball hoops for most of COVID. But they are back now!​I shot around for half an hour today and my shot is still kinda ok! Hopefully I’ll start to play some pickup games in the next couple weeks.<img src="https://cdn.lu.ma/editor-images/o6/d4ec2873-60b2-48b8-aeb8-39ee45608b4b"/>​<h2>​Coming Up</h2>​This week I have a few exciting things:<ol><li>​my 2nd vaccine shot (with the Danqings)</li><li>​a marathon in Marin on the day after my shot (hopefully I’m in shape to run it)</li><li>​a short Luma retreat (to spend more time on the planning I mentioned above)</li></ol>​In terms of work, we’re planning on launching a member directory feature for Luma — a place where you can see bios and information about other members of your community. ​If you know of an online community that might benefit from a member directory, please let me know!</div>

Weekly Update 2021-04-04

In a time long forgotten, there was a web framework called Meteor. Meteor let you share code across your frontend and backend. This code sharing let you avoid the hassle of creating a REST API and keeping your client and server in sync. With Meteor, you would write Meteor.call and the Meteor service would translate that into an HTTP request.While this was very cool, Meteor didn’t end up catching on. My theory was that Meteor was too early and trying to do too much. And also React took over frontend dev. Ever since Meteor fell from grace, I have wanted a better way of sharing code between the frontend and server but haven’t found a good way of doing it. While code sharing is still a dream, at Luma we’ve figured out how to share types across the client and server. Defining the InterfaceWe create Typescript interfaces for GET and POST requests. Each interface looks like this:shared/get-api.tsexport type GetApi = {
 "/event/get": {
 request:{
 event_api_id: string;
 };
 response: {
 event: {api_id: string; name: string};
 };
 };
 ...
}Then on the client we can do something like this:client/api.ts// This will use the GetApi to type the function inputs and outputs
public get = async <
 Path extends keyof GetApi,
 Request extends GetApi[Path]["request"],
 Response extends GetApi[Path]["response"]
>(
 path: Path,
 params?: Request,
 customHeaders?: any
): Promise<Response> => {
 return await this.request("get", path, params, customHeaders);
};

// this.request just wraps axios and does our auth logic for us

// using this
const {event} = api.get('/event/get', {event_api_id});Then to type our Koa server requests:server/router.tsexport const routerGet = <
 Path extends keyof GetApi,
 Request extends GetApi[Path]["request"],
 Response extends GetApi[Path]["response"]
>(
 path: Path,
 routeFunction: (
 ctx: RouterContext<Record<string, unknown>, IRouterContext>,
 requestData: StringifyValues<Request>
 ) => Promise<Response>
) => {
 router.get(path, async (ctx) => {
 const requestData = ctx.query;
 ctx.body = await routeFunction(
 ctx,
 requestData
 );
 });
};

// Then to define an actual route we do something like
routerGet("/event/get", async (ctx, {event_api_id}) => {
 // event_api_id is typed from GetAPI
 // and what we return is also typed! so we will get a type error if we don't return the right thing

 return {event: {api_id: 'victor', name: 'pontis'}}
})(Btw, Koa is very similar to Express. So I expect this kind of thing would transfer over well to Express.)Sharing Code with Yarn WorkspacesWe structure our code in a mono-repo. Each folder corresponds to a different service — client/, server/, etc. But we store shared code in a special place, the shared/ folder.The shared/ folder is compiled and shared across all of the services. We use the shared/ folder to store common utility functions and types.This is where we define the API interface between the client and server.Yarn Workspaces lets us easily share code without publishing to npm. We simply run yarn in our Dockerfile and the shared code just works.BenefitsWe get the benefits of type-checks across all of our API requests. When you change the signature of an API route, you can quickly see every other invocation that is affected. This typechecking extends to the React components and server routes. We can even type our React components by API responses — const Component(props: PostApi[‘/route-name’][‘response’])You can use code completion in the editor to auto-complete API routes. If you type in LumaClient.post you will see a list of suggested routes and payloads. The typescript API serves as a tacit documentation for the API — one place to view all routes and their expected types.DrawbacksIn order to keep the types simple, I banned all parametrized API routes. So we can’t do POST /events/<evt_api_id>/update. Instead, we do POST /events/update and pass in a the event ID. This doesn’t seem to hold us back but it feels non-standard. Every time you add a new request, you have the added overhead of defining the API interface. I think this is clearly worth it, but there is definitely extra overhead. The nice-ness of the API typing has let us be inconsistent with route and argument names. For example, we may do /event/list but also do /newsletters/list.

<div class="mirror-content" data-reactroot=""><img src="https://cdn.lu.ma/editor-images/zq/57fb998c-0f2e-49e5-9d81-cc93de52f2fe"/>​In a time long forgotten, there was a web framework called <a href="https://www.meteor.com/" target="_blank" rel="nofollow noopener">Meteor</a>. Meteor let you share code across your frontend and backend. ​This code sharing let you avoid the hassle of creating a REST API and keeping your client and server in sync. With Meteor, you would write <code>Meteor.call</code> and the Meteor service would translate that into an HTTP request.​While this was very cool, Meteor didn’t end up catching on. My theory was that Meteor was too early and trying to do too much. And also React took over frontend dev. ​Ever since Meteor fell from grace, I have wanted a better way of sharing code between the frontend and server but haven’t found a good way of doing it. ​While code sharing is still a dream, at Luma we’ve figured out how to share types across the client and server. ​<h2>​Defining the Interface</h2>​We create Typescript interfaces for GET and POST requests. Each interface looks like this:​<code>shared/get-api.ts</code><pre>​export type GetApi = {
 &quot;/event/get&quot;: {
 request:{
 event_api_id: string;
 };
 response: {
 event: {api_id: string; name: string};
 };
 };
 ...
}</pre>​Then on the client we can do something like this:​<code>client/api.ts</code><pre>​// This will use the GetApi to type the function inputs and outputs
public get = async &lt;
 Path extends keyof GetApi,
 Request extends GetApi[Path][&quot;request&quot;],
 Response extends GetApi[Path][&quot;response&quot;]
&gt;(
 path: Path,
 params?: Request,
 customHeaders?: any
): Promise&lt;Response&gt; =&gt; {
 return await this.request(&quot;get&quot;, path, params, customHeaders);
};

// this.request just wraps axios and does our auth logic for us

// using this
const {event} = api.get(&#x27;/event/get&#x27;, {event_api_id});</pre>​Then to type our Koa server requests:​<code>server/router.ts</code><pre>​export const routerGet = &lt;
 Path extends keyof GetApi,
 Request extends GetApi[Path][&quot;request&quot;],
 Response extends GetApi[Path][&quot;response&quot;]
&gt;(
 path: Path,
 routeFunction: (
 ctx: RouterContext&lt;Record&lt;string, unknown&gt;, IRouterContext&gt;,
 requestData: StringifyValues&lt;Request&gt;
 ) =&gt; Promise&lt;Response&gt;
) =&gt; {
 router.get(path, async (ctx) =&gt; {
 const requestData = ctx.query;
 ctx.body = await routeFunction(
 ctx,
 requestData
 );
 });
};

// Then to define an actual route we do something like
routerGet(&quot;/event/get&quot;, async (ctx, {event_api_id}) =&gt; {
 // event_api_id is typed from GetAPI
 // and what we return is also typed! so we will get a type error if we don&#x27;t return the right thing

 return {event: {api_id: &#x27;victor&#x27;, name: &#x27;pontis&#x27;}}
})</pre>​(Btw, Koa is very similar to Express. So I expect this kind of thing would transfer over well to Express.)​<h2>​Sharing Code with Yarn Workspaces</h2>​We structure our code in a mono-repo. Each folder corresponds to a different service — <code>client/</code>, <code>server/</code>, etc. But we store shared code in a special place, the <code>shared/</code> folder.​The <code>shared/</code> folder is compiled and shared across all of the services. We use the <code>shared/</code> folder to store common utility functions and types.​This is where we define the API interface between the client and server.​Yarn Workspaces lets us easily share code without publishing to npm. We simply run <code>yarn</code> in our <code>Dockerfile </code>and the shared code just works.​<h2>​Benefits</h2><ul><li>​We get the benefits of type-checks across all of our API requests. When you change the signature of an API route, you can quickly see every other invocation that is affected. </li><li>​This typechecking extends to the React components and server routes. We can even type our React components by API responses — <code>const Component(props: PostApi[‘/route-name’][‘response’])</code></li><li>​You can use code completion in the editor to auto-complete API routes. If you type in <code>LumaClient.post</code> you will see a list of suggested routes and payloads. </li><li>​The typescript API serves as a tacit documentation for the API — one place to view all routes and their expected types.</li></ul>​<h2>​Drawbacks</h2><ul><li>​In order to keep the types simple, I banned all parametrized API routes. So we can’t do <code>POST</code> <code>/events/&lt;evt_api_id&gt;/update</code>. Instead, we do <code>POST /events/update</code> and pass in a the event ID. This doesn’t seem to hold us back but it feels non-standard. </li><li>​Every time you add a new request, you have the added overhead of defining the API interface. I think this is clearly worth it, but there is definitely extra overhead. </li><li>​The nice-ness of the API typing has let us be inconsistent with route and argument names. For example, we may do <code>/event/list</code> but also do <code>/newsletters/list</code>.</li></ul>​​</div>

We made our development 10x faster by typing our API interface

Hi again! My last update was actually this Wednesday. So here I am again in your inbox…wow, it looks like you have a lot of unread messages here 👆👇.What’s new…Luma — Newsletter ImprovementsI made a bunch more improvements to newsletters in Luma (which was used to compose this email). There are a bunch more improvements coming this week that will make Luma a viable Substack replacement for a large group of people. Reading — Origin of WealthI started this book a couple months ago and then basically stopped reading any books. So I’m getting back on the train. The book argues that Traditional Economics (what they teach in textbooks) has made ridiculous assumptions in order to simplify the math. These assumptions — perfectly rational actors, perfect information, etc — lead to incorrect predictions. The author is beginning to explain how a group of economists and physicists came together at the Santa Fe Institute to try to come up with a more sophisticated theory of analytics.  I liked learning about how some early Economists admired the works of Poincare and Newton and pulled in concepts from physics and mechanics to create the first formulations of supply and demand. A Year of LumaWe started Luma a year ago. I missed the exact year mark from the first commit. So I’ll have to write a recap on April 3rd when we’ll be a year from this article.

<div class="mirror-content" data-reactroot="">​Hi again! My last update was actually this Wednesday. So here I am again in your inbox…wow, it looks like you have a lot of unread messages here 👆👇.​What’s new…​​Luma — Newsletter Improvements​I made a bunch more improvements to newsletters in Luma (which was used to compose this email). There are a bunch more improvements coming this week that will make Luma a viable Substack replacement for a large group of people. ​​Reading — <a href="https://www.amazon.com/Origin-Wealth-Evolution-Complexity-Economics/dp/157851777X" target="_blank" rel="nofollow noopener">Origin of Wealth</a>​I started this book a couple months ago and then basically stopped reading any books. So I’m getting back on the train. ​The book argues that Traditional Economics (what they teach in textbooks) has made ridiculous assumptions in order to simplify the math. These assumptions — perfectly rational actors, perfect information, etc — lead to incorrect predictions. ​The author is beginning to explain how a group of economists and physicists came together at the Santa Fe Institute to try to come up with a more sophisticated theory of analytics. ​I liked learning about how some early Economists admired the works of Poincare and Newton and pulled in concepts from physics and mechanics to create the first formulations of supply and demand. ​​A Year of Luma​We started Luma a year ago. I missed the exact year mark from the first commit. So I’ll have to write a recap on April 3rd when we’ll be a year from <a href="https://techcrunch.com/2020/04/03/how-to-prevent-zoombombing/" target="_blank" rel="nofollow noopener">this article</a>.​</div>

Weekly Update — 2020-03-28

So I’m not sure if I can call this a weekly update anymore... My last weekly update was on December 7th. A lot has happened since then!One of the more exciting things is that I’m sending this from Luma! I’ve been working on improving our newsletter capabilities and they’ve reached a point where I can use Luma instead of Substack 😃.OK, let’s see what else has happened lately…Vaccinations!In the past month or so a ton of people I know have gotten vaccinated. Both my mom and my dad. Friends at higher risk.And I got vaccinated too —A friend told me on Thursday night that there were a bunch of open appointments in Sacramento that they wanted to fill. So we booked an appointment and drove 2 hours to get a vaccine the next morning. Luma — Improved Dashboard + Audience ViewWe recently rolled out a new Luma dashboard. With that comes with a much more powerful audience view. (That’s actually where I’m sending this newsletter from!)I’m Learning GreekRight now I’m going thru the process of applying to get Greek citizenship. (My mom’s parents are from Greece and immigrated to America in the early 1900s.)So as part of the process I’m trying to learn Greek. I just had my first tutoring lesson last weekend and I’ve been watching episodes of Sto Para Pente to gain more exposure myself in the language. My biggest learning so far is that the way we teach langauge in school is sooo broken. I spent 6 years learning Spanish via vocab cards, grammar rules, and conjugations and I still barely speak Spanish. Instead we should be focusing on understanding media and communication. Then language learning will happen as a byproduct. I’m trying to learn Greek by understanding a TV show, not by learning the grammar rules of the Greek language. 

<div class="mirror-content" data-reactroot="">​So I’m not sure if I can call this a weekly update anymore... My last weekly update was on <a href="https://victorp.substack.com/p/weekly-update-2020-12-07" target="_blank" rel="nofollow noopener">December 7th</a>. A lot has happened since then!​One of the more exciting things is that I’m sending this from Luma! I’ve been working on improving our newsletter capabilities and they’ve reached a point where I can use Luma instead of Substack 😃.​OK, let’s see what else has happened lately…​​Vaccinations!​In the past month or so a ton of people I know have gotten vaccinated. Both my mom and my dad. Friends at higher risk.​And I got vaccinated too —<img src="https://cdn.lu.ma/editor-images/av/b021f459-0993-45f7-81f6-8ad57e450b81"/>​A friend told me on Thursday night that there were a bunch of open appointments in Sacramento that they wanted to fill. So we booked an appointment and drove 2 hours to get a vaccine the next morning. ​​Luma — Improved Dashboard + Audience View<img src="https://cdn.lu.ma/editor-images/j1/51d53c30-1584-4de4-8041-b495c66a6889"/>​We recently rolled out a new Luma dashboard. With that comes with a much more powerful audience view. (That’s actually where I’m sending this newsletter from!)​​I’m Learning Greek​Right now I’m going thru the process of applying to get Greek citizenship. (My mom’s parents are from Greece and immigrated to America in the early 1900s.)​So as part of the process I’m trying to learn Greek. I just had my first tutoring lesson last weekend and I’ve been watching episodes of <a href="https://en.wikipedia.org/wiki/Sto_Para_Pente" target="_blank" rel="nofollow noopener">Sto Para Pente</a> to gain more exposure myself in the language. ​My biggest learning so far is that the way we teach langauge in school is sooo broken. I spent 6 years learning Spanish via vocab cards, grammar rules, and conjugations and I still barely speak Spanish. ​Instead we should be focusing on understanding media and communication. Then language learning will happen as a byproduct. I’m trying to learn Greek by understanding a TV show, not by learning the grammar rules of the Greek language. </div>