Check the work
Every post links to the files and logs behind it. This page collects those entry points in one place.
No result here asks to be taken on trust.
Hacking my own Twitter for AI agents
The app, the attack scripts, and every row from the ten-attacker drill.
- Open the section in the postwhat each file proves
- Download the bundle (.zip)code, test scripts, raw results
- README.mdwhich file backs each claim
Training my first model
The split script, the training script, the console output, and the trainer's per-pass record.
- Open the section in the postthe two rebuild steps
- Download the bundle (.zip)scripts, output, hashes
- README.mdwhich file backs each claim
AI Does Not Want to Kill Us
The paper's retained responses, reconstructed traces, and its Python checker.
- Open the section in the postthe exact verify command
- Verification package (.zip)includes verify.py and checksums
- Evidence tableevery retained response