I gave my own product an MCP server: 20 tools, two scopes, and the default that misled a real analysisTL;DR Connecting your AI to somebody else's MCP server takes one line of config. Writing the server for your own product is where the actual decisions live, and almost none of them are about the protJul 28 · 13 min readcodemeetai.hashnode.dev
AI for builders · 04: context engineering, simplifiedTL;DR. A prompt is the ask; context is the ground it stands on: the goal, the real material, the constraints, and the boundary of what to ignore. When the AI gives you generic or wrong output, the refJul 20 · 9 min readcodemeetai.hashnode.dev
AI for builders · 03: 90% of Claude Cowork isn't code, and that's the half I stopped doing by handTL;DR. Chat makes you drive every step; Cowork flips it, you describe the outcome and it works across your files and apps and hands back the finished thing. Anthropic says more than 90% of Claude CowoJul 13 · 7 min readcodemeetai.hashnode.dev
AI for builders · 02: the memory folder that caught a bad deploy before I shipped it Tags: ai, productivity, claude, developer-tools, indiehackers TL;DR.: Every new session your AI starts from zero: no memory of you, your repo, or the thing it got wrong yesterday. The fix is not a feJul 10 · 10 min readcodemeetai.hashnode.dev
How I'm running loop engineering with Fable 5"Originally published at codemeetai.substack.com. TL;DR I pointed Fable 5 at my own React Native / Expo SDK, read only. It found 11 real bugs, each cited to a file and a line, six more than the two otJul 8 · 14 min readcodemeetai.hashnode.dev
