index Everything, newest first

25 posts

2026

  1. New AI Benchmark: Ludowars
  2. Halloween special: Skeletons
  3. Getting to the bottom of things with Claude Code
  4. The end of human software construction more radical than thought
  5. The Only Moat is Money
  6. Frontier lab test and eval is fundamentally unsafe
  7. How to think like a product manager
  8. First software 3.5 model?
  9. AI vs Math vs P vs NP
  10. The problem is still that the AIs are kinda bad, not that they are good
  11. The worst people with the best AI
  12. The end of AAAI
  13. Even the APIs are below the API
  14. Is software still a career?
  15. The unreasonable effectiveness of stochastic parrots
  16. Coding is over. But the specs still need to be written
  17. Some bad takes from the internet
  18. Flashback to simpler times
  19. Opening a Time Capsule from 1996
  20. Genius 5 - A browser-based 'analog' synth

2025

  1. Triviality tests and the sycophancy problem
  2. Software 2.5, Software 3.0 and Software 3.5
  3. Thinking? What thinking?
  4. I made a casual math game
  5. Back Back Back