AI Learnings
Shorter takes on tools and workflows. Less "here's a guide" and more "here's a thing that actually worked."
Peak AI Frenzy
You AI everything. You automate all the things. Then you find the checkpoints, realize you overbuilt, and come back down to using less. The comedown is the healthy part.
The only AI advice I actually give people
Use Claude for every task for two weeks, and refuse to route around a single blocker. That's the whole method. The gains compound faster than you'd believe.
Five things in my AI preferences that actually changed my output
Preferences are not a setup step you complete. They're a living system you tune. Start with the heckle rule, which took one line and changed everything.
A task that reports success and writes nothing
Eight weeks of watching my own scheduled AI tasks fail. The one that hid the longest never crashed once.
The handoff doc is the real product
I rebuilt my Research / Plan / Implement skills around one idea: each phase's deliverable is a 200-line baton, not the pile of work behind it.
The folder system that made AI actually useful
Most people re-explain their entire job to AI every single conversation. The fix isn't better prompts. It's better folders.
Claude Cowork Mode is the real deal
Been using it to build this site and manage my project roadmap. It's like pair programming with someone who actually reads the docs.
Five tools. That's the whole stack.
I've tried dozens of AI tools. Five survived. Each one does something the others can't.