Run
Started the morning with an 8-mile run — new distance PR. Heart rate felt controlled the whole way through, though I’m pretty confident the watch is overstating by 20–30 bpm. Showed at the end that I can push the pace when I need to, even if the last stretch was mostly downhill. Felt like an incredible way to start the day.
Earned hotpot tonight.
Agent auditability
Random idea: a good coding agent should be able to explain why it made certain decisions, and those explanations should be auditable. Not sure how big an issue this is in my existing workflow, but it feels like something that should be reliable — being able to trace back agent reasoning would be genuinely useful.
Codex overnight
Codex ran for 8 hours yesterday. Pretty incredible to see. The loop was running integration tests after every change — each test taking over an hour — but it just kept going. The fact that it natively captures Claude Code skills like /loop is powerful. Bought the $100/mo Codex plan.
Game dev
Making real progress on the game and had fun playing against Claude yesterday. The idea I keep coming back to: a marketplace of AI games — including something Survivor-style where you play against AI agents — feels like it could genuinely go viral. Further down the line, there might be agent swarm optimizations worth exploring here to learn inference engineering and dogfood thunder-compute infra at the same time. But that’s way further out.
Investing
Still trying to think through how the forum should look. No conclusions yet.