The two-person YC startup says its cloud reasoning agent beat GPT-6.2 Astra and Dream-1 on bimanual tasks it had not seen before, and opened a preview API.
By [Ryan Merket](https://runtimewire.com/author/ryan-merket)
· Published
Waddle Labs says its first robot-use agent, Waddle-1, scored 81 in Robocurve PBC (@robocurve) evaluations, ahead of GPT-6.2 Astra at 65 and Dream-1 at 59.
The San Francisco startup announced Waddle-1 in a thread on X on November 20th, saying the system controls bimanual robots on unseen tasks zero-shot. Robocurve's public leaderboard separately lists GPT-6.2 Astra at 65.7% and Dream-1 at 59.5% on its RoboDojo-RC Tier 1 benchmark, consistent with the rounded comparison figures in the announcement.…