What just changed
Published on October 8, 2026 by OpenAI News: LegalOn reduced its estimated daily Codex costs by 65% while maintaining development speed. The published summary names the method: the team matched Astra, Sol and Luna to suitable tasks and managed budgets strategically.
No new feature here: it is an operations lesson.
What the case shows
The summary names Astra, Sol and Luna without describing each one's role. So nothing on this basis says which handles which tasks, or how they perform. What is established is the method: the 65% saving goes hand in hand with per-task matching and budget management.
- Published figure: -65% estimated daily Codex costs, per OpenAI.
- Constraint held: development speed is maintained, per the same summary.
- Named levers: matching Astra, Sol and Luna to tasks, and budget management.
Where the case stays silent
The summary publishes no benchmark, no pricing and no definition of "estimated cost". It does not say how much work went to each of the three names either. That matters: the 65% is LegalOn's own result, and nothing suggests it transfers unchanged to another codebase.
Three levers to activate this week
- Inventory the tasks handed to Codex and their estimated daily cost.
- Test a different pairing on one category of simple tasks, comparing speed and quality.
- Set a visible budget, since budget management is one of the levers named for LegalOn.
Which task handed to Codex could switch pairing without anyone noticing?
If you follow the latest AI technology, I publish a deep dive every day on frontier models, hardware, robotics, automations and AI-generated music. 👉 Get the next one straight in your inbox — sign-up takes ten seconds.