MJ

After watching numerous agents over and over again either: - pull in dozens of different packages and stick them altogether, or - rewrite schema validation from scratch Remix 3 being a single package that takes you very far on its own might be the biggest selling point

@BrooksLybrand 🤘

Launching Rusl today — collaborative tooling for schema-driven design. Publish schemas. Review changes. Pin versions. Install them like dependencies. rusl.com

@boshen_c @arshadyaseeen Kudos to you for calling it out 👍 learning from each other is powerful. Open source ftw 🙌 Oxc is fantastic btw. Our team has been getting a ton of use out of it.

I want to recognize that Yuku from @arshadyaseeen is faster than Oxc. It uses data-oriented design everywhere, which is truly amazing. I’m glad there is something faster for us to learn from. Oxc needs really big breaking changes to get to the same level of performance, but I'll consider it. github.com/yuku-toolchain…

Since you’re asking for feedback, a few high value feature requests: - Have you seen how Claude code on the desktop helps you file “chips” for work that it discovers along the way? Very convenient way to follow up on side tasks. - my codex threads need to have some visibility into what the other threads are doing. I’m constantly multitasking. If thread A is waiting for issue B to come to a resolution, it’s good to know that thread C is working on B and when it’s done. - codex desktop does not know when I’m creating PRs. The agent is using the gh tool to create and manage PRs, but the interface has no clue. Compare to Claude code desktop that is able to easily manage multiple PRs on a single thread at once. Codex requires me to click the button to create a pr, taking me out of the normal chat loop. - viewing diffs is completely broken on codex desktop when I switch between threads. Just doesn’t work at all. It stays on the diff for the last thread I was looking at. This might sound like all bad bad bad but I promise you I love the work you guys are doing. The only reason I give feedback is because I want the product to be even better! 🫡

I work with the best team in the world. we use our own product all day and all night and will not (and do not) sleep until it’s perfect. and it’ll never be perfect keep sending us your thoughts– we’re here for you. thank you all ❤️

@theo Claude code is the best harness. It’s not even close.

@sregg You'd be surprised

@mjackson I bet your employer has a very different take 😅

Prediction: synthetic native app architectures will not survive much longer. Hybrid web/native apps and stuff like react native and flutter were always an attempt at making native apps easier to build for non-native builders. They’re much less appealing in an AI coding world.

More useful than many benchmarks: GPT 5.6 Sol just improved tensor parallel prefill numbers preserving numerical correctness in DwarfStar after many rounds of Fable optimizations. Does not mean it is better, bit it is a strong signal IMHO.

@thdxr That whole post reads like catharsis for some feelings that have apparently been building for years now.

this part made me lol i was ready for some big reveal on how jarred was trying to corrupt the honorable Zig Foundation in pursuit of self gain and it was that he asked for an LSP

Are you familiar with the Ruby spec project? Years ago a bunch of folks wanted to try implementing Ruby on different platforms because C Ruby (the original) was too slow for them. So they tried all kinds of things. But as they got into the implementation, they found that matz had made a bunch of assumptions about how Ruby was supposed to work that were effectively baked into the C Ruby implementation. So Ruby spec was invented to try and distill all that knowledge into a runnable spec. It was the spec that allowed alternative implementations to be built that could run eg Ruby on Rails apps. Without it, they would’ve all had many different quirks that made behavior unpredictable.

@mjackson I don't agree with that but fair enough. But i am saying like if we just make the only impl the spec we don't even have to think about a spec vs implementations. The spec just becomes the implementation.

@Shreyassanthu77 Deriving the spec from the implementation is more difficult and lossy than the inverse. If I describe my app in plain English, I can implement it however I want. The implementation is secondary.

@mjackson But why not just have a singular source of truth that is also the only implementation instead of a spec + 2 or more implementations? just more parts that could get out of sync and stuff

@Shreyassanthu77 Business logic moves from code into markdown files/specs/plain English so any agent knows the requirements without deriving them from a specific implementation.

@mjackson hmm but duplicated logic is bad tho no? even if llms can do it easily i still think smth like kotlin multi platform is still useful

@jarredsumner Legend. Thanks for sharing 🖤

@jaredpalmer @jarredsumner @turborepo Shhhhhh! Don’t say it out loud. I don’t want them to start charging more for the models!

The Bun rewrite was dirt cheap even at Fable pricing. For comparison, it took 3-5 engineers (about half the core team) about ~14 months to rewrite @Turborepo in Rust between 2023 and 2024. When you consider the opportunity cost on top of eng salaries, it actually cost several millions of dollars.

@schrockn Same. I’ve been using Fable a ton, and I was blown away at how quick and efficient Sol was to work with today. I think I had just gotten used to very slow progress.

Pure vibes, but Sol seems to have *way* more token throughput than Fable.

@mitsuhiko @jarredsumner It’s developer envy. What Jarred did is nothing short of remarkable.

As someone who maybe is not the biggest fan of Bun, I think what @jarredsumner is doing is pretty fucking awesome and I can confirm that he's great to hang out with and just very friendly. The responses to his blog post by some people are uncalled for.

@schrockn @jarredsumner Good grief. That post is absurd.

I have never actively wanted a programming language to fail but thanks to Andrew Kelley's scurrilous, outrageous, classless post on @jarredsumner 's bun rewrite, I got there. Port every project from Zig to Rust.

@ahejlsberg @typescript Congrats! We've been using the tsgo native preview for months now and it's been fantastic 👍 Excited to finally get onto 7.0 stable!

Huge milestone for our team today: TypeScript 7 is now generally available--a native port that runs 10x faster. @typescript devblogs.microsoft.com/typescript/ann…

📣 The moment is here. 📣 TypeScript 7 is officially released! 7️⃣ devblogs.microsoft.com/typescript/ann…