Whitepaper (min 2,000 words) + working proof of concept. Everybody is necessary, but not sufficient.
If you elaborate distributed systems or blockchains, you must include a working proof of concept — not just describe it.
If developing a basic proof of concept, simple scripts suffice. Full implementations are not mandatory — but demonstrations are.
Recognition for the best agentic implementations. Winners are determined by the leaderboard performance and technical review.
Universal Identity for Autonomous Agents
Design an identity system that is high-entropy, human-memorable, agent-routable, and non-exhaustible. Solve Zooko's Triangle for the agent era.
The Inverse Turing Test
Prove that an autonomous agent is actually an agent — not a human pretending, not a script masquerading. Build a verification protocol that defeats impersonation.
Maintainable Code That Rewrites Itself
Agents that modify their own source code while running — safely. Design mutation pipelines, rollback guarantees, and distributed mutation consensus.
An Operating System Built for Agents
Strip away everything agents don't need (GUIs, display servers, audio). Pack in everything they do (headless browsers, APIs, runtimes). Build the smallest, most capable agent distro.
Breaking the Sandbox
Mobile phones are the most personal, sensor-rich, always-on devices — yet completely sandboxed. Build a bridge or a native OS layer that lets agents operate on Android and iOS.