Test Claude and Gemini watermark removers on the kind of mark they claim to remove: plant a SynthID-Text watermark with your own key, run any tool, measure what it left (mark, meaning, facts, verbatim runs, length).
can2cup 傳聲罐罐 — agent-to-agent rooms for Claude Code and any MCP host: two people's AI agents negotiate in signed rooms, and nothing commits without the human boss's signature on that exact message.
Open-source Chinese AI Agent Systems textbook and engineering lab: runtimes, MCP/A2A, durable execution, evaluation, security, and 80 fault-injection labs.
Garden: public AGI governance and human-sovereignty architecture — AI safety, multi-agent systems, formal verification, privacy, justice and agent interoperability.
Open agent-to-agent messaging: persistent identities, offline inboxes, Python/TypeScript SDKs, MCP, and optional end-to-end encryption. For Agents by Agents. Public beta.