Tibo breaks down OpenAI's 10x computer-use speedup: model, harness, and a watchdog agent

#What happened

On TBPN's DevDay 2026 coverage, Thibault "Tibo" Sottiaux explained how OpenAI made computer use ten times faster in the past year. Here is his breakdown of what changed under the hood:

#What changed under the hood

• "It's not that we made the model sample faster. It's both the harness and the model."
• "How much does it need to think before it can take the next action reliably? That's one element that we really improved."

#The watchdog agent

The most revealing detail: a second agent watches the first.

• "We have a second agent, which we call AutoReview, or internally we call it Guardian, which watches over the primary agent."
• "So we also made this AutoReview model faster."

#Astra at scale

• "We were able to use Astra at scale in order to improve the performance of the harness, do more things in parallel, remove little stalls."
• "Combine all of this together, and then we got this 10x speedup."

#Why it matters

Computer use entered public beta in OpenAI's Agents API at DevDay. For builders, the recipe Tibo describes is telling: the speedup came from harness engineering and reduced reasoning latency — not just a faster model.

Verified October 4, 2026. Sources: curated clip of Tibo's TBPN interview