Goszczu

Example z.ai/blog/glm-4.7
These percentages are so abstract that is hard to evaluate is this better in everyday usage than SOTA.

What strikes me that all of AI releases show benchmark percentages but none is giving how many tokens used or how long they were running. 80% != 80%. Details matter

If you automate this with Linear and Slack. You can shorten time to PR dramatically. Most importantly you don’t commit into any additional AI subscriptions while keeping control of the whole execution env, what MCPs can be used, what agent can and can’t do.

I’m astonished how easy was to set up opencode in GH Actions. One command a few clicks and everything works like charm. Now I can chat with big pickle in my PRs

I’m trying to set up a productive ai workflow in nvim. Currently I’m using sidekick with opencode. I was sitting on Avante for a very long time but amount of configuration needed to switch between models was too much. The only thing I’m missing is in place ai editing rn. I guess I have to write it

Supabase | tonik Part spaceflight, part internet spectacle - we turned Supabase Launch Week 14 into a fully immersive, retro-futuristic experience.

Super happy to see @supabase.com LW14 live. It was quite a journey since we at @tonik.com had only three weeks for everything. You can check out what we’ve built.

Whenever I have to find a repository on GH a part of me dies. Why is the UX of it so bad? :(

Emperor new clothes book cover, with text overlay “BTC edition”

Why I can’t get rid of the feeling it’s accurate?

Despite all this, we got a working plugin! WebSocket-powered, CRDT-backed, and even supporting browser editors.

I’ll share the code once it’s cleaned up and a few bugs are ironed out. Stay tuned! ⚡️ #neovim #rustlang #lua #crdt #hacking

3️⃣ Rust async gotchas? As Rust newcomers (more async Rust newcomers), we underestimated how much time we'd spend debugging errors like "cannot move XYZ because Foo doesn't implement Send", where Foo was some deeply nested type. On top of that, most of LLMs didn't know how to fix these issues too.

2️⃣ Nvim API quirks? Turns out, there's no way to track byte-level changes—only full-line diffs on remote API. Not ideal for real-time collaboration. We had to create a custom event type and send it via a socket from Lua to Rust.

1️⃣ Lua ↔ Rust communication? After some trial and error, we ended up spawning a separate process and using NeoVim’s remote API for better isolation. We thought about using nvim-oxi or mlua, but after a few tries to set up it correctly, we backtracked to bulletproof rmp.

We used Lua + Rust, with CRDT (via the yrs library) under the hood. Nice API, solid Node & Rust support. But, of course, we hit some roadblocks:

🚀 A week ago, I joined Tonik’s hackathon with my brother, and in just 24h, we built a working PoC for a NeoVim "Live Share" plugin—despite having almost zero experience with nvim plugins. It was super fun.

Mistral released le chat but I want le avante in my neovim.
Anyone tried to set it up with custom provider? How painful it is?

Joking if not obvious :D

Link’s broken

Ok, @tonik.com is here! It’s official 💙

Photos app logo

Kindly submitting team application

Shower thought: Factorio for developers is the same as ETS for truck drivers

Some Playwright types now reported as `any` after upgrading to 5.7.2 · Issue #60619 · microsoft/TypeScript 🔎 Search Terms I manually scrolled through all issues reported since the release of 5.7.2. 🕗 Version & Regression Information This changed between versions 5.6.3 and 5.7.2 ⏯ Playground Link https:/...

Ok, this seems related github.com/microsoft/Ty... 🤔

testLoggedInWebapp.describe("As User", () => { testLoggedInWebapp.describe("Homepage", () => { testLoggedInWebapp( "Shows logged in user", // Error, loggedInWebapp: any, user: any async ({ loggedInWebapp, user }) => { expect( await loggedInWebapp.isLoggedIn(user), ).toBe(true); }, ); }); });

testLoggedInWebapp.describe("As User", () => { testLoggedInWebapp.describe("Homepage", () => { testLoggedInWebapp( "Shows logged in user", // No error, PlaywrightTestArgs & ... correctly infered async ({ loggedInWebapp, user }) => { // Add empty line expect( await loggedInWebapp.isLoggedIn(user), ).toBe(true); }, ); }); });

I hit a super weird case. TS fails to infer Playwright types, but when I add one empty line to the test body, everything starts to work again.

Check Avante (my personal pick) or Code companion

Ok, my new hobby. Going through anvaka.github.io/map-of-github/ and finding hidden gems

Story: Writing Scripts with Go Story: Writing Scripts with Go. GitHub Gist: instantly share code, notes, and snippets.

I wrote a little bit of Go. It was refreshing how fast its code can be executed, so I thought about writing some shell scripts with it. Then found this gist.github.com/posener/73ff...
I guess `go run ...` and setting aliases is the way.

Does it start with „exit” and ends with „animations”?

Probably there are a lot of other approaches that I'm now aware of.

For now I feel that special documentation focused on mental model of project modules and project architecture can improve "0-shot" answers. Supplying this with framework docs should yield better results. Going to test that.

Devin style learning via PRs and comments. This is just a broader case of the History+RAG approach. It can work but requires time and effort. I think it's good for big projects but not for small/medium ones.

History + RAG, this could work and I think some of the IDEs/tools are doing it already. The biggest issue is that needs to learn gradually based on feedback and the amount of code it seen. That's why it feels like some context is missing.

A simple AI.md file committed in a repo can be added automatically to each prompt. The question is how it's different from docs or ARCHITECTURE.md. Maybe someone has already tried it.