      Tomasz Tunguz Blog · 2025-10-09
      · 238d
    

AI Engineering Promotion: From Intern to Senior Engineer in One Year

Tomasz Tunguz analyzes OpenAI's Codex advancement from junior to senior engineer status, examining adoption metrics and optimal collaboration patterns. He presents architect-implementer systems and closed feedback loops as key design patterns for human-AI engineering collaboration, with OpenAI achieving 7 hours of autonomous execution and 15K lines of code refactored.

5 metrics· Cited 0× in the knowledge base ·Open source ↗

Metrics in this report

Autonomous Execution Duration Record

7hours

maximum achieved

OpenAI architect-implementer pattern

Code Refactored in Autonomous Execution

15000lines of code

single execution record

OpenAI architect-implementer pattern

Daily Codex Usage Rate

92%

among technical staff

Engineering teams using Codex

Pull Request Generation Increase

72%

relative increase

AI users vs non-users

Token Consumption in Autonomous Execution

150000000tokens

in single execution

OpenAI code refactoring task