PUBLIC RESEARCH · SERIES 01
Field notes from the transfer gap.
Our baseline claims are published before interviews so practitioners can challenge them. After each five conversations, we will revise the evidence, preserve material disagreements, and record what changed.
MethodThese first three notes synthesize public documentation, issue reports, and research—not private interviews. Sources and inference are labeled.
September 8, 2026 · 6 min read
Five reasons robot policies fail to transfer
A baseline synthesis of public LeRobot reports and cross-embodiment research, published before our first interview cycle so practitioners can challenge it.September 8, 2026 · 5 min read
What is missing from most robot-policy repositories
A practical inventory of the evidence needed to decide whether reproducing a policy is feasible before purchasing hardware or spending a week integrating it.September 8, 2026 · 7 min read
The minimum reproducibility record for manipulation research
A deliberately small proposed record that can live beside an existing GitHub or Hugging Face repository without forcing anyone onto a new platform.September 8, 2026 · 4 min read
What the first robot-skill validator cannot infer
Release notes from turning the proposed reproducibility record into a working local inspector—and the evidence automation must not pretend to know.CORRECT THE RECORD