OpenAI engineer tests GPT-5.3-Codex running about 25 hours continuously to build a design tool from scratch
Original title:Run long horizon tasks with Codex
AI summary
An OpenAI developer blog documented a long-horizon experiment: using GPT-5.3-Codex with Extra High reasoning to build a design tool from scratch in an empty repository, running uninterrupted for about 25 hours, consuming about 13M tokens and generating about 30k lines of code.
Key Takeaways
- An OpenAI developer blog documented a long-horizon experiment: using GPT-5.
- 3-Codex with Extra High reasoning to build a design tool from scratch in an empty…
- consuming about 13M tokens and generating about 30k lines of code.
Quick Facts
- Published
- Updated
- Source
- OpenAI Developers:Blog(网页)
- Category
- Tutorials & Tips
- AI score
- 67
作者用一次25小时的实测复盘,给出长时程 Codex 任务的文件栈和验证方法,可迁移到自己跑长任务。
Related entities
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: OpenAI engineer tests GPT-5.3-Codex running about 25 hours continuously to build a design tool from scratch. February 23, 2026. https://new.nat6.net/en/news/u0owjbxlr4b7n2ulykpxzo1o1/