I Planted One Bad Instruction in Three Places in a Repo. Copilot Read Them All. It Never Bit.

In-browser RAG with WebLLM and WebGPU, a hero animation that fires with every token, and the four mistakes I fixed on the way to Lighthouse 100.

I ran Liquid AI's LFM3-1.7B in the browser with WebGPU and built Grading Buddy, an offline study app that answers only from notes I've checked.

Every client wants something different, and AI makes it cheap to build. I built a zero-dependency browser sandbox so those customisations stay out of

If you've watched an AI coding agent build a to-do app in two minutes, you've probably asked yourself...

Why streamText in @ai-sdk/google dropped Gemini File Search and RAG citations, how the merged fix dedupes sources by identity, and how to upgrade to 4.0.118.

Quando estava construindo o Downshift, precisei entender melhor como o Claude Code gastava contexto e...

On December 1, 2026 at 9:00 AM CET, remove.bg — the site that turned "one-click background removal"...

I analysed 43 hours of my agents’ run history. The model took only 16% of it. The rest was tests,...

A few weeks ago I wrote about LLM Explained, a free, open-source course of 43 interactive pages that...

hikyaku is an open-source Go message relay: one YAML file, retries and fallbacks at every step, a dead-letter queue you can replay, and an AI router.