This is a report about a number, so let us start with the number.
Between December 4 and December 11, contributors wearing Concern gloves captured 61 million hours of dexterous manipulation data. Verified, quality-scored, consented, provenance-tracked. In one week.
For scale: the combined published datasets of the entire academic robotics field—every grasp dataset, every teleoperation corpus, every motion-capture study from 1965 to the founding of the Concern—total approximately 40 million hours, and most of it is lab data, which is to say most of it is a thousand variations of picking up a standardized block.
The field's first sixty years: 40 million hours of blocks. Last week: 61 million hours of surgery, carpentry, cooking, hair braiding, cello, drywall, sushi, sutures, and one contributor in Reykjavik who repairs antique typewriters and whose data the licensing agents describe as "unreasonably valuable."
We want to be precise about why this happened, because "growth" is not an explanation, it is a description.
First: the Registry itself is compounding. More contributors joined in November than in the entire first year. Every new contributor adds capture capacity that never sleeps, because hands never stop.
Second: wear time per contributor is rising. Twelve months ago the median contributor wore gloves 2.1 hours per day. Today it is 6.4 hours. Gen 2 gloves disappear on the hand; people forget to take them off. Forgetting is the product working.
Third—and this is the part the roboticists said was impossible—data quality is rising with volume. The naive assumption was that scaling from hundreds of contributors to hundreds of thousands would flood the corpus with noise. The opposite happened. The dexterity manifold identified in the 847-person study becomes better resolved with every hand added. Rare motions that appeared once and were flagged as anomalies now appear a thousand times and resolve into structure. The corpus is not getting bigger and noisier. It is getting bigger and sharper.
At current trajectory, the corpus doubles every nine weeks. We have checked this figure four times because it does not look like a real number. It is a real number.
A researcher at a robotics lab we will not name—one of the labs that spent a decade on simulated grasping—wrote to us after the Q3 report. We publish the message with permission:
"I spent nine years generating synthetic grasps. You generated more useful data during this email."
The lab has applied for a licensing agreement. So have forty others this month.
The curve is vertical. The hands keep moving. ☜