Backend freshness system

Scheduled agents keep the catalog alive.

This page explains the research automation behind the hub without pretending every crawler result is truth. New findings go into a review queue before they become public claims.

Current Records 0 seeded entries
Public Posts 0 human-approved only
Review Policy 2x source evidence minimum
Cadence Daily discovery sweep

Schedule

Agent Workflows

Design only
Daily Discovery Sweep

Search GitHub topics, GitHub code, ROS Discourse, Reddit, arXiv, YouTube, Hackaday, OSHWA, newsletters, and robotics blogs.

Every 2-3 days Verification + Deduplication

Confirm licenses, source files, CAD/BOM/docs, duplicate lineage, dead links, abandoned repos, and misleading openness claims.

Weekly Classification + Scoring

Apply lane taxonomy, readiness scoring, velocity scoring, home-robot fit, and missing-evidence tags.

Weekly Benchmark Watcher

Track claims, papers, demos, build logs, telemetry, and third-party reproductions for challenge protocols.

Approved only Publisher

Draft site updates, newsletter notes, and X/Twitter posts after a human confirms the project entry and wording.

Crawler targets

Where the agents look

GitHub robotics ROS Discourse arXiv robotics Hackaday robots OSHWA Awesome robots

Publishing Guardrails

Required
  1. Do not call a project open source unless license evidence is present.
  2. Separate software-open from hardware-open, CAD-open, and BOM-open.
  3. Label scraped findings as candidates until manually reviewed.
  4. Keep public updates factual, sourced, and modest.
  5. Send uncertain claims to the community AI as questions, not assertions.