Harry Nguyen

Hope people aren’t tired of the “just use Postgres” meme yet (surely not @BenjDicken), but Postgres can be your filesystem too.   Today I’m launching sql-fs: a persistent bash sandbox over HTTP and MCP, backed by Postgres.   Most agent sandboxes still spin up VMs: cold start time, per-minute billing, and costs that climb when you run lots of sessions. I went a different way. just-bash is the virtual bash runtime. I built the distributed and persistent layer around it, which is a Postgres-backed virtual FS with consistency across replicas and in-process caching, so sandboxes are cheap to create, quick to resume, and durable across process restarts.   What you get: • Filesystem state in your Postgres (RLS per sandbox) • Content-addressable blobs (dedup across sandboxes) • 3 layers of caching and locking for distributed exec • Path and content caches so stat/read don’t hammer the DB every time   Repo: github.com/Hazzng/sql-fs   I also wrote a longer blog post (see reply below) on how it works in depth. Have a read if you’re curious about the distributed design to achieve strong consistency :)   PS: thanks to @cramforce for the amazing just-bash!

you know what we all should do

@p0u4aa @nmazzy i don’t think @jarredsumner would agree lol

@nmazzy zig 4 life

A little something I've been working on. The coding agent I actually want: N.O.V.A Made for parallel work. Version control primitives built into the harness. Review the code as the agent generates. I also fine-tuned a BERT model for auto-reviewing bash tool calls.

When building a product, ask yourself "will this choice seem like a good idea in 5 years?" The "I'll fix it later" mentality it good in *some* cases, but the long-term vision keep products reliable, functional, scalable. I don't mean pre-mature optimization. Don't shard your database when you have 10 users. Don't optimize global pageloads when your entire customer base is US. Rather, choices like: - The framework you build with (and how) - The database at the core of the infrastructure - The dc / cloud you're building on - The organization of your monolith These are very important decisions, and ones that get harder to change with growth.

@p0u4aa damn huge

@BenjDicken @samlambert i literally just finished all of ur series :) next one when??

@hazzngn @samlambert sounds like a great stream topic

@samlambert would love a detailed analysis from PScale’s folks on OpenAI’s postgres setup for scaling to serve 800M users

@samlambert Totally agree the single primary node is defs a bottleneck in that setup. Im so keen already to read more about Neki’s detailed architecture

@hazzngn OpenAI proved postgres can scale pretty far on read workloads. They did not prove postgres can scale write workloads. the posts own solution to write scaling is to move those workloads to cosmo. not much of a lesson here. Luckily Neki fixes all this.

@samlambert @PlanetScale @BenjDicken cant be more true!! learned a ton from pscale

If you are early in your career and want to get a broad overview of various systems fundamentals you should read everything we post on the @PlanetScale blog and follow @BenjDicken.

This is number 4: a unified abstraction for all agent harnesses. Write the agent once. Run it on any harness without vendor lock-in vercel.com/changelog/prog… Huge thanks to @felixarntz for pushing this over the line!

@p0u4aa @badlogicgames damnn

This took way too long to figure out and hack together but step 1 is done (it moves). Easily the most fun i’ve had building in a while. Permission to join the shitty robot club @badlogicgames

@thsottiaux I built a distributed virtual filesystem backed by postgres that achieve strong consistency for as many agents working in the same env as we want with 3 layers of locking and caching for optimal performance (its fast!) x.com/hazzngn/status…

I have a new kind of big button that I can press for Codex. Over the next 100 days, we will select one person per day who does impressive or incredibly useful work with Codex and give them 10X usage limits for a month to see what they can do with it. First one tomorrow.

I don't watch many talks but Andrew Kelley's "A Practical Guide to Applying Data Oriented Design" had a big impact on me. First watched it right around when I started working on Ruff.

@nmazzy @jarredsumner tough choice

@hazzngn @jarredsumner what about gta6

i do actually believe mythos will be rolled out before @jarredsumner’s blog post now

@cramforce @BenjDicken haha thanks big fan of vercel-lab

@hazzngn @BenjDicken Looks awesome

Hope people aren’t tired of the “just use Postgres” meme yet (surely not @BenjDicken), but Postgres can be your filesystem too.   Today I’m launching sql-fs: a persistent bash sandbox over HTTP and MCP, backed by Postgres.   Most agent sandboxes still spin up VMs: cold start time, per-minute billing, and costs that climb when you run lots of sessions. I went a different way. just-bash is the virtual bash runtime. I built the distributed and persistent layer around it, which is a Postgres-backed virtual FS with consistency across replicas and in-process caching, so sandboxes are cheap to create, quick to resume, and durable across process restarts.   What you get: • Filesystem state in your Postgres (RLS per sandbox) • Content-addressable blobs (dedup across sandboxes) • 3 layers of caching and locking for distributed exec • Path and content caches so stat/read don’t hammer the DB every time   Repo: github.com/Hazzng/sql-fs   I also wrote a longer blog post (see reply below) on how it works in depth. Have a read if you’re curious about the distributed design to achieve strong consistency :)   PS: thanks to @cramforce for the amazing just-bash!

also thanks alot to @p0u4aa for contributing and finding bugs :)

Here’s the blog post harryng.dev/posts/persiste…

First video up! I'm a research student from Australia, and I want to build a little circle of people learning this stuff seriously and pushing each other to get better. Sharing my journey as I go, starting with visiting DETR. If that's you, join in :) youtube.com/watch?v=RpgSdr…