Notes

AI lets the CEO prototype on a Sunday and the CPO ship a flow by Tuesday. The question nobody has answered yet: what does the team do with it?

No prompts. Just hooks. Tasks auto-sync to Apple Reminders as a byproduct of running Claude Code on top of an Obsidian vault.

The things worth sharing are usually the ones you built for yourself first.

Chat history resets. Documents don't. The principle behind building accountability into your Claude Code setup.

Claude has a habit of saying 'I'll remember that' without writing anything. Built a hook that blocks the response until it does.

For 30 years, technical execution was scarce and taste was abundant. AI flipped that. The people who spent the last five years developing genuine aesthetic judgment now have an enormous, unearned advantage.

45+ APIs now accept x402 payments. Agents can access all of them autonomously. No keys, no OAuth, no subscriptions. This is what the machine-to-machine economy looks like.

In 1991, HTTP shipped with a 402 status code reserved for payments. It never got used. Agents might be what finally makes it real.

I want to say please and thank you to Claude, but I don't want to waste tokens on it. So I shipped a one-line CLAUDE.md rule. Token optimization or AI survival insurance? Yes.

Sharp tools beat Swiss Army knives.

I built an MCP server that gives any AI assistant structured World Cup data. 18 tools, zero API keys, works offline. Here's why bundled data beats API calls.

Opus is the safe default when you're building fast. But a three-line tier table in my system prompt taught the model to flag when it's overkill.

1930s Hollywood hired chaos agents called Wildies to pitch insane ideas on set. Every industry needs one -- and keeps trying to optimize them out of existence.

In an era of rapid platform and protocol churn, the smartest move is building systems that survive collapse -- not ones that predict the future.

AI agents are getting better at acting on our behalf, but they have nowhere to land. The web isn't built for them yet.

The best generative art projects aren't about aesthetics -- they're about designing systems that produce unexpected outcomes within intentional constraints.