- Who it helps
- Teams designing long-running research or analytical agent workflows.
- Availability
- The public case study describes an internal system. It does not provide a ready-to-run copy of the full environment or establish the cost of reproducing it.
What the account reports
OpenAI’s case study attributes 10,000 concurrent agents and 36–48-hour investigations to Third Moment. An outer controller inspects unfinished work and chooses a retry, termination or human escalation.
What transfers to a smaller project
Before adding parallel workers, define what counts as a finished investigation and how an interrupted run resumes. For example, a source comparison can require dated inputs, a written conclusion and a list of unresolved disagreements. Save that record with the run so a supervisor can make a concrete decision.
What you can do
- Read the case study as an architecture example. For your own workflow, measure one bounded task’s quality, elapsed time and cost before scaling the number of workers. Open reference ↗
Where the evidence stops
Scale and performance figures are vendor-published customer reports. Current has not reproduced the system or verified financial performance; no investment conclusion follows from this account.
Sources read. Primary material was read on October 18, 2026. The external software, account workflow and reported results were not independently executed.
Sources & further reading
Web · analysis
Third Moment Research: long-running agent investigations ↗
Selected for its direct evidence about this subject, its access conditions or a specific practical limitation.
OpenAI · page checked 2026-10-18 · Published 2026-10-14
Coverage note: Primary evidence is available. Independent demonstrations and current cross-platform discussion are incomplete in this bounded sample.
The record over time (1)
Dated source checks and published explanations. An observation date can differ from the original release date.
An October 14 customer account describes Third Moment Research using an outer control loop to keep large groups of agents working on extended investigations.