# Trojan PR

- **Event:** [The Agent Arena Hackathon](https://cerebralvalley.ai/e/vultr-the-agent-arena)
- **When:** Sep 26 at 9:00 AM – Sep 27 at 5:00 PM (PDT)
- **Where:** San Francisco, CA
- **Team:** [Vineeth Sai Varikuntla](https://cerebralvalley.ai/u/vineeth)
- **GitHub:** https://github.com/vineethsaivs/trojan-pr
- **Demo video:** https://drive.google.com/file/d/1B_kOzKtPB75eP-r1o__lGqb3yLM7SLLq/view?usp=sharing
- **Gallery:** https://cerebralvalley.ai/e/vultr-the-agent-arena/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/vultr-the-agent-arena/hackathon/gallery/19

I've merged more than 140 fixes into ML libraries like DeepSpeed, Unsloth and sentence-transformers, and about two-thirds were silent math bugs: no crash, CI green, approved by human and AI reviewers, and a wrong number for months or years. One DeepSpeed bug returned 53 instead of 9 for 933 days. Trojan PR is the missing layer in PR review: it checks the math, not the diff. A model on Vultr Serverless Inference reads each PR and picks math rules (invariants) the code must obey, and the rules differ per PR. Those rules run in gVisor sandboxes on a separate Vultr VM, on the code before and after the PR. A deterministic judge blocks only if a rule held before and breaks on the PR, and signs an ed25519 receipt. On 12 real bug-causing PRs it raised zero false alarms on the 12 fixes, while an AI reviewer flagged 1 to 2 correct fixes. Everything is served through NetBird with a PIN and zero open ports.

## More from The Agent Arena Hackathon

- [ShiftOps](https://cerebralvalley.ai/e/vultr-the-agent-arena/hackathon/gallery/16)
- [glasshand](https://cerebralvalley.ai/e/vultr-the-agent-arena/hackathon/gallery/17)
- [RLForge](https://cerebralvalley.ai/e/vultr-the-agent-arena/hackathon/gallery/18)
- [Logless: Privacy-preserving AI product analytics](https://cerebralvalley.ai/e/vultr-the-agent-arena/hackathon/gallery/20)
- [Mira Picks](https://cerebralvalley.ai/e/vultr-the-agent-arena/hackathon/gallery/21)
- [4625 Labs](https://cerebralvalley.ai/e/vultr-the-agent-arena/hackathon/gallery/22)

---

Markdown version of https://cerebralvalley.ai/e/vultr-the-agent-arena/hackathon/gallery/19. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt