Tomek Czajecki

@konstipaulus @birch_js Oh yes for LTR Latin I think HarfBuzz is an overkill. But for anything more I would definitely use it now (I thought I am the last person to give up on doing it from scratch but I looked at the problem space and thought do I really want to spend MONTHS researching this)

@birch_js @tchayen Harfbuzz is heavy and less performant compared to text-shaper. Although I haven’t tested anything but latin ltr so far, it may not be a full replacement depending on the requirements

Implemented the Slug Algorithm in WebGPU. It uses a pure TypeScript text shaping pipeline to convert Inter.ttf into Bézier curves completely without Harfbuzz or WASM. Tysm for dedicating the patent to the public domain. Repo below

@satya164 2. So the statement is technically (and grammatically) correct

@tchayen How many people have asked you?

People ask me what's my secret for not running out of tokens. It's simple. My Claude finds the slowest CI command that it can run locally and runs it 5 times with random combinations of | tail -n and head that tell it nothing until it gives up and runs the full thing.

@plahteenlahti I told claude to find itself some tool for this and stop asking me to test things manually and it came up with this github.com/joshuayoes/ios… so far it seems to be working reasonably

@tchayen I'm the mcp for that (but replying to follow answer, didn't even know you could do that)

What MCP do you all use for giving claude hands and eyes for use in iOS simulator?

Introducing Expo Agent Build truly native iOS and Android apps from a prompt. Anything from React to SwiftUI to Jetpack Compose. Compile and deploy for Apple, Android, and the web right from the browser!

Exactly one year ago (10 mar 2025), Dario Amodei: "I think we will be there in 3-6 months, where AI is writing 90% of the code. And then, in 12 months, we may be in a world where AI is writing essentially all of the code." This turned out to be... too darn accurate.

@SaadNajmi We switched some of our repositories too. The difference is crazy. Especially if you still have some eslint rules left that couldn’t be migrated.

We switched a couple of repos over to TSGo, oxfmt and oxlint. We’re seeing massive speed improvements, building speed is near unrecognizable

@VadimNotJustDev Rare W for Apple

Apple banned my app 🙃 How is your monday?

Not sure if I managed to mute all agentic coding grifters or they had to return the stacks of mac minis before the deadline

Last week I accidentally went to sleep at 10pm and woke up at 6am rested like never before but my body didn’t fall for it and we are back to 4am scrolling I guess

You know you are not getting anywhere today with your problem when chatgpt quotes your stackoverflow question from 2y ago as the main source

@av3nxr openclaw is great and use it too - believing you need to buy mac minis and run simultaneous agents on different hardware to replace employees (which is complete bs) means that you have lost touch with reality

No amount of LARP-ing gets anywhere close in efficiency to just talking to the LLM There's probably a normal distribution meme about this

@hirbod @aleqsio Ok maybe not “just”. Setting up Tailscale and Termius wasn’t that fast haha. Maybe better to go to sleep…

@hirbod @aleqsio Just install claude code, tmux and finish setting it up from your bed

Remember: latest Age of Empires 4 DLC is out since November and it’s just $20 dollars you don’t need a Mac Mini to feel like you are playing an RTS game

@hirbod Hell yeah. Anyone who tried setting up anything resembling normal web browsing on a VPN headless chrome or inside a Raspberry Pi knows how much it just won’t work (@aleqsio 🥲). Feels like the life of my OpenClaw is so much harder because it was born poor 😭

Sent while my OpenClaw on a VPS is updating Astro docs for a product it built earlier

@sebastienlorber IMO side projects are perfect for that. I made the most progress in building understanding and intuition around using LLMs in parallel agentic workflows just building several throw-away projects. It’s how you can push the limits without being afraid of making mess

I feel I'm falling behind on all these AI things I'd like to force myself to use agents more for real open-source work, and not just throw-away side-projects What's a good way to get started? Is it even a good idea to add an AGENTS md to an OSS repo?