brbrown @brbrown

@jaykreps @confluentinc @ShaunMClowes Jay your mark on people, culture and tech are huge. Really excited to see what comes next for you!

A form of Conway's law: high quality internal tools lead to high quality external tools. Here's Hector showing our internal dev tool we use to build our deck icon sets, including publishing the icons.

Hector Simpson @dizzyup

7 days ago

Making dev tools for our dev tools. 'Super Icon Studio' is how I've been designing the icons for our 'Deck' icon groups - it covers design finish across light/dark modes, layout compositions, animations, an icon catalogue, and ships as a highly customizable Swift package ready

14 5 450 113K 161

This one is actually really good.

Phil Eaton @eatonphil

a week ago

Clay.com's AI Writing Policy. This is well done.

2 34 205 26K 212

Seeing an unprecedented influx of pull requests from accounts submitting dozens of PRs across many completely unrelated repos in a short period of time. I don't know what their motivation is for doing so, but what I do know is that we don't have the review capacity (nor the interest) for dealing with them, it's eating up maintainer time we'd rather spend elsewhere. So I sat down to clarify the guidelines around AI-based contributions to the Hardwood project, following our mantra "Built with AI, not built by AI". OK 👍: - Use all the tools you want for creating PRs, but you must understand what the changes are and you must be able to stand behind them in a review. Verify a) that the original issue was sensible to begin with, b) that the generated code solves that issue, c) that it does so in line with the project's standards for correctness, completeness and performance. Not OK 👎: - Point an LLM to an issue from the tracker, accept its output without consideration, and open a pull request - Pull requests opened by autonomous agents/bots, without any human in the loop - Pull requests from accounts mass-submitting large numbers of PRs across many unrelated repos (there was one account opening 252 PRs across 98 different repos on a single day) To be clear, this isn't about tooling. I use AI every day, and much of Hardwood is built with it. It's about whether there's a person who can stand behind a change. I'm sure these rules will continue to evolve over time, but we gotta start somewhere. #Hardwood #OpenSource #LLM #AI

So the serving layer for this distributed graph at Netflix queries an 8-billion node, 150 billion edge graph in under 100ms? Cool deep-dive here, including the important role of gRPC ... netflixtechblog.com/how-and-why-ne…

The main cause for the continued debate around reviewing generated code: vastly different consequences when shipped code fails. If a bug causes your side project to crash, you'll arrive at a different position than if user data, money, or worse is at stake.

Had Eric Lippert on the podcast to talk about his new book Fantastic Adventures in Data Structures and Algorithms - now it's out! On sale in early access. hubs.la/Q03Q9Q9y0 @ManningBooks

📺 "It's built with AI, not built by AI." Had the chance to chat about #Hardwood with one of my favorite podcast hosts, the inimitable @tlberglund. We touched on a range of subjects: why I think the JVM ecosystem was in need of a new library for Apache Parquet, some of the tricks Hardwood employs for parsing Parquet fast, how we're using AI for building the project (hint: it's not vibe-coding), what's ahead on the roadmap, and more. Thanks for having me, this was fun! 👉 youtube.com/watch?v=9Ov0Cn…

Announcing Discovery Loop! I am very excited to announce that, along with my longtime friends and collaborators @Sanjay_Ghemawat, @OriolVinyalsML and @quocleix, we are founding Discovery Loop (@DiscoLoopAI), a Public Benefit Corporation whose mission is to automate machine learning, science, and engineering to accelerate discoveries and progress. The four of us have worked together for 14 to 30 years, and have helped build some of the world’s most used products, infrastructure and AI models, and we’re excited to turn our attention to this ambitious endeavor. ♾ Learn more at: discoveryloop.com

keyboard_arrow_left Previous keyboard_arrow_right Next

Probably the biggest startup announcement of the year (decade?) in terms of the calibre of founders. If I had to name the top 10 engineers currently or previously at Google in terms of impact it would be the two cofounders (Larry and Sergey) and all 4 of these folks:

Jeff Dean @JeffDean

2 weeks ago

Announcing Discovery Loop! I am very excited to announce that, along with my longtime friends and collaborators @Sanjay_Ghemawat, @OriolVinyalsML and @quocleix, we are founding Discovery Loop (@DiscoLoopAI), a Public Benefit Corporation whose mission is to automate machine

keyboard_arrow_left Previous keyboard_arrow_right Next

897 2K 22K 6.6M 6K

Kyo rewrote Postgres and MySQL wire protocols so you never have to see JDBC again 🥳 RC6 is out. Write once, run anywhere, turbocharged 🚀🚀🚀 kyo-sql gives you raw SQL plus a typed DSL that mirrors SQL syntax, statements suspend fibers instead of blocking threads, and one set of sources compiles for JVM, JS, Native, Wasm, and GraalVM Under it all is kyo-net and kyo-ffi: one C transport and TLS stack over io_uring, epoll, kqueue, NIO, Node, and BoringSSL. Yes, your Node.js process now runs io_uring and BoringSSL 😱 Heads up: JDK 25 is now required for the main modules Bonus: kyo-ai now runs on your Claude Code or Codex subscription instead of per-token billing 💸

In the last 18 months or so my terminal use has exploded. From being a place where I mostly navigated the file system, ran a server and did some very light nvim use, to my entire workspace, connected to multiple machines, where I spend most of my day. This is the starting point we're using for designing the @superlogical multiplexer, and we're borrowing ideas from another app that blurs the local and remote, long lived sessions and throwaway tabs: the web browser. And the first thing I wanted to borrow: favicons, for your terminal apps. We call it the Deck icon (like a deck of cards), since it needs to represent your own layout of (potentially) multiple running processes.

Deck icons: a dynamically changing set of icons based on what's running in your multiplexer. You've seen SuperSplit (early iterations) before, but this is also a small showcase of SuperTabs, our custom native tab kit. And yes, vertical tabs already work too.

almonk @almonk

2 weeks ago

In the last 18 months or so my terminal use has exploded. From being a place where I mostly navigated the file system, ran a server and did some very light nvim use, to my entire workspace, connected to multiple machines, where I spend most of my day. This is the starting point

133 77 2K 301K 1K

I lead migration efforts at @PlanetScale and I’d say 90% of the most painful migrations we’ve seen have been due to queue workloads in Postgres. At scale it consumes system resources disproportionate to data size. generally requiring near constant auto-vacuums and a battle for contention with other queries. The referenced post dances around this scale challenge. So while it works.. for a bit.. it often leads to premature db scaling challenges when other techniques/archs wouldn't. wrong tool for the job, nearly always, IMO

Qian Li @qianl_cs

3 weeks ago

Postgres actually scales if you use it right.

9 10 117 163K 63

📢 A decade after I published Quill, kyo-sql is landing with a fresh new approach! 🙌 ✅ First-class support for both raw SQL and typed queries ✅ Two-way interop: raw SQL in typed queries and typed queries in raw SQL 🤩 ✅ Support for JVM, Native, Wasm, JS on top of Kyo's new high-perf async network stack ✅ Pluggable drivers, MySQL and Postgres included, with DB-specific features ✅ Compile-time query generation without friction via inlining 🙌🏼 Something that has been great is how Kyo's Record has enabled much simpler and ergonomic designs. It's a cornerstone of a few modules so far: fully typed http routes (including filters!), seamless state handling in durable workflows, and now typed SQL queries with semantics close to relational algebra, reducing friction. Yes, flexible structural typing with rich type-level transformations is a reality in Scala 3 with Kyo! 🧙🏼‍♂️ I realized a long time ago that expressing SQL as collection/monadic operations like Quill did has always been a dead end. This new work corrects the record and mirrors SQL syntax in the typed DSL instead This module gets us much closer to 1.0 in time with the upcoming Scala 3.9 LTS. It's been a long journey, and we're finally close to the finish line!!! 🔥🔥🔥

The Ghostty non-profit has signed our first long-term paid contributor via a 12-month agreement with @vancluever. He's a top-quality engineer, has been part of the Ghostty community for years, and has consistently demonstrated an extremely high quality and judgement bar. He is the creator and maintainer of z2d, the vector graphics library that Ghostty uses. He has contributed upstream to the Zig compiler and Arocc C compiler dozens of times. The way Ghostty contributor contracts work is that there is no stated or mandated area of focus. I trust Chris to work on worthwhile tasks, and Chris has demonstrated he does so far, so there's no need to push him one way or the other. He chooses what to work on and describes his work to the non-profit (so it's all aligned and legal to compensate). As I spend more time focused on libghostty, especially the libghostty needs of @superlogical, I want to ensure that the broader Ghostty project continues to be staffed and healthy. We'll be signing more contributor contracts soon!

I fear I am about to be the recipient of Baby's First Restraining Order after I gift this to my friend's child.

Let's talk a bit about what makes the Superlogical multiplexer different architecturally from traditional terminal multiplexers! There's a lot (a LOT) more coming, but wanted to share a little bit about the terminal-specific part compared to tmux and zellij.

I've started a new company: @superlogical! We're going to begin by building a terminal multiplexer. The entire vision is much larger, but the multiplexer is the foundation. Sign up for the newsletter to get beta access and devlogs (product updates only I promise). superlogical.com

Still figuring out my way with AI, like everyone else. But one red line I’m sure about: don’t let AI speak for you without disclosure. Sharing raw AI output can actually be fine if you say so. For instance, I’ll occasionally send review findings prefixed with “via Claude Code, apply what makes sense.” What’s just rude is passing off machine text, which you may not even have read yourself, as your own words.