All Posts

I replayed 800 AI agent runs inside a sandbox that records what the Linux kernel sees, and asked a small model to watch them with and without that evidence. A pre-registered test couldn't show it helped – here's what I learnt along the way.

Benchmarking DeepSeek V4 Flash and GLM 5.3 Flash on a MacBook Pro – why parameter counts only predict the start, why the built-in draft model slows things down, and what it costs to keep your RAM back

Why the most important skill with AI coding agents isn't prompting or curating context – it's designing how they verify their own work

How to sandbox coding agents while maintaining development flexibility - the ideal balance between security and convenience

How do I mix local LLMs with cloud ones and code gen with continuous learning

Outright rejection of gen AI won't get us the best outcomes

I hereby renounce years of suffering with native extension compilations breaking

How Lisbon has changed in the last 30 years

How to set up another drive for a home setup

Part two of the series Building applications using embeddings vector search and Large Language Models