HN Explorer

Show HN: AI vs. AI – code and reviews only count if they survive an attack (github.com)

1 points by lolu1032 ↗ HN
Pantheon is a technique for generating each subagent, generating code with different solutions, breaking it and writing code that ultimately survives. For Pantheon-X, GPT 5.5 is broken, and GPT is broken and surviving is the last skill to write code written by Claudette. In the case of Pancheon Gap, it is a technology to continue to check whether the actual review is true and to derive surviving children.

I'd appreciate it if you could actually use it and give me feedback through issues

1 comment

[ 3.0 ms ] story [ 11.2 ms ] thread
ericmaciver
The version that boosted improvement for me: make the reviewer a different model than the one that wrote the code, and prompt it to be adversarial: refute rather than green light. A launch-blocker fix that compiled and passed every test still had 3 real bugs - the adversarial cross-vendor pass was the only thing that caught them. Same-model self-review mostly rubber-stamps. That being said, I usually run same-vendor in parallel though, just for instant consensus on the major issues, but I direct fixes based on the adversarial model. The "survive an attack" part is key.