Philosophy
Kill Ugly Raw Source, AI Rewrites
This is Halo's founding philosophy, and the true meaning behind its name.
Originally, the phrase referred to using AI to refactor bad code. Today, its scope has expanded far beyond code — it points to everything in the digital world that is raw, inefficient, and requires repeated human intervention.
We believe: anything that AI can handle automatically should no longer consume human attention.
How Halo Is Built
Halo is a living experiment in AI productivity. Its first version was created in half a day, and from that point on, 100% of the code has been generated by AI — without ever opening another IDE.
This is not a gimmick. It is a serious validation of an engineering approach:
Phase 1 (Human Core Value): Spend 2–4 hours in deep discussion with AI about design, architecture, and technology stack — down to the directory structure — producing a comprehensive development plan of several thousand words. Architectural decisions are the most irreplaceable human contribution.
Phase 2 (AI Execution): Hand the development plan to AI and grab a coffee.
Phase 3 (Review): About half an hour later, dozens to hundreds of files are ready,
npm run buildpackages it, and a cross-platform desktop application launches directly.
Halo's architecture incorporates VS Code's layered design philosophy while abstracting Agent pluggability, enabling a shared rendering layer across Web, Windows, macOS, and Linux.
Traditionally this might require a team working for months or years. With top AI models, one person can accomplish the same in a single weekend. This is an industrial-revolution-level leap in efficiency — just as a single driver can operate a locomotive to haul tens of thousands of tons of freight.
Our Position
Halo is not a commercial product. It is an open-source project driven by personal and community interest, dedicated to exploring more possibilities of AI Agents in spare time.
We believe the computing platform of the future should not merely offer Copilot tools, but should provide a standardized runtime environment where AI agents can persistently and autonomously take over complex business workflows. This should be open, community-built, and not monopolized by commercial companies.
About the DHP Protocol
Based on the above philosophy, we proposed DHP (Digital Human Protocol) — an open AI digital human distribution protocol.
Developers write a spec.yaml, submit it to the DHP Registry, and users worldwide can install and use your digital human with a single click.
See → DHP Protocol