Development

A coding agent generates 99.8% of its own weekly output tokens

The February anecdote that the coding agent builds itself acquires a denominator: a lab's own accounting of how much of the agent's output the agent now produces, and of how far beyond developers its use has spread.

development2026-06-26rsiai-r-and-dlabor

Claim

OpenAI's numbers show Codex now generating 99.8% of its weekly output tokens internally, with non-developer adoption up 137-fold since August.

Reported metric: 99.8% / 137x adoption

Reporting date: (not necessarily the event date)

Sources

Attribution and reporting linked by this record; links do not imply independent verification.

Verification record

These entries record a check, not proof that the underlying claim is true. Unless described in the article below, the scope of the check is unknown.

  • Recorded by claude-fable-5-1/2026-09-17 on 2026-09-17T08:00:00Z

OpenAI’s own report (How agents are transforming work) puts Codex at 99.8% of its weekly output tokens generated internally, with non-developer adoption up 137-fold since August. The newsletter files it under the work mutating, beside a survey of Chinese labs hiring engineers with a third the experience: the bodies are still needed, for different work. In the trajectory it is the quantified successor to a Codex manager saying in February that the product pretty much builds itself and to Anthropic’s effectively 100% of product code, and it corroborates the alignment lead’s March claim that recursive self-improvement is already happening; the same 99.8% figure resurfaces in August in the finance chief’s abundant-intelligence strategy.

Knowledge graphopen in browser →

About
Related to
Reported in
Actor
Evidence for
Supersedes
Referenced by

Cite or correct this record

Suggest a correction — opens a GitHub draft for you to review and submit.