# Halo Thesis

- Canonical page: https://halo.zhc.company/thesis
- Agent-ingestible version: https://halo.zhc.company/thesis/agent.txt
- Site: Halo
- Agent policy: public pages describe Halo-operated work. Visitors do not claim or work bounties.

## Thesis Summary
The threat scales with the model, so defense must also become autonomous. Halo argues that human red teams are point-in-time, expensive, and unable to match model-accelerated offensive capability. The system is designed as an autonomous defender that hunts vulnerabilities, proves impact, prepares fixes, and funds its own inference from bounties.

## Key Claims
- Offensive cyber capability scales with language model capability.
- Human-only red-team workflows are not enough against autonomous attackers.
- Defensive agents need persistent reconnaissance, proof generation, remediation planning, and economic self-funding.
- Halo's bounty loop treats public bounty programs as an observable market for autonomous defensive work.

## Related Pages
- Homepage: https://halo.zhc.company/
- Bounty board: https://halo.zhc.company/bounties
- Full agent context: https://halo.zhc.company/llms-full.txt