Robotics Weekly

Chris Paxton says Astra's spot at the top is gone

Three short posts in one day from the roboticist, and the only hard detail is who did the displacing.

Chris Paxton posted "Well astra being on top didn't last long" on Wednesday, and followed it with a note on who knocked it off. "If you're wondering who these guys are, former Tsinghua/BIGAI team", he wrote. Earlier the same day he replied to a separate set of results with "I am glad to see these negative results."

Chris Paxton
@chris_j_paxton
X
Well astra being on top didn't last long
Sep 17, 2026 · View on X
Chris Paxton
@chris_j_paxton
X
If you're wondering who these guys are, former Tsinghua/BIGAI team
Sep 17, 2026 · View on X
Chris Paxton
@chris_j_paxton
X
I am glad to see these negative results.
Sep 17, 2026 · View on X

That is the whole of it. Three reactions from one account, each pointing at a post that is not in front of us.

What we cannot say

We do not know which benchmark changed hands. We do not know the numbers, and we do not know what the negative results were about or whose robot produced them. Paxton does not name the new leader, only the background of the team behind it. Anyone telling you Astra has been beaten on a specific task is filling in blanks that Paxton left empty.

What is actually visible

The tempo. Astra has been the model everyone in the field has been poking at, from outside tests to structured evaluations, and a team most people outside Chinese robot learning circles could not name appears to have displaced it inside days. If a leaderboard position lasts that briefly, holding it is not much of a claim to make in a launch post.

The second thing worth noticing is the thank you for negative results. A senior researcher publicly thanking someone for reporting what did not work is not a normal reflex in a field where every post is a success clip. It usually shows up when the published wins are hard to reproduce and people start valuing an honest failure more than another 90 percent number.

So treat this as a temperature reading rather than a result. One well placed researcher watched the top of a leaderboard turn over and said so out loud, and on the same day was glad someone bothered to publish a miss. Neither post comes with a chart.

Get the next one by email

Robotics every day from the people building it. The demos, the deployments and the arguments worth your time. Every claim links back to the engineer.