Morning scan creates one timestamped log for each Agent across current papers and open-source projects.
RAG Agent
Retrieval, grounding, context and citation.
Open research →PAT / Learning Research & Development
Learn twice.
Improve carefully.
Five Foundation Agents scan papers and open-source projects at 08:30 and 18:30 Asia/Shanghai. Findings become timestamped logs, not automatic production changes.
Memory Agent
Memory, storage, forgetting and hardware.
Open research →Tool Use Agent
Tools, Skills, MCP and safe execution.
Open research →Planning Agent
Harness, constraints, loops and recovery.
Open research →Evaluation Agent
Benchmarks, validity and regression.
Open research →Twice-daily learning logs.
Evening scan repeats discovery and records a separate learning outcome. Open private GitHub logs →
Public-safe research only. Private prompts, credentials, personal data
and Domain Team secrets never enter this surface. Every candidate
remains UNREVIEWED until Human and Benchmark gates pass.