이미지: X — 프론티어랩
Summary
- OpenAI's developer account announced a feature for Codex and ChatGPT that understands recent work context
- Opting into 'Computer History' lets users pick up where they left off and receive suggestions for skills or scheduled tasks based on recurring work patterns
- It can be seen as an extension of the ChatGPT Work-Codex sync feature unveiled on August 11
- 발표
- OpenAI Developers, 2026년 8월 13일
- 기능명
- Computer History (옵트인 방식)
- 적용 대상
- Codex, ChatGPT
- 핵심 동작
- 최근 작업 맥락 파악 → 이어서 작업, 반복 작업 패턴에 스킬·예약 작업 제안
- 관련 선행 발표
- 2026년 8월 11일, ChatGPT Work-Codex 프로젝트·대화·스킬·플러그인 동기화 (챗GPT 데스크톱 앱)
What happened
OpenAI's developer account announced on August 13 that it added a new feature to Codex and ChatGPT. It's called "Computer History." Once opted in — meaning users must turn it on themselves — ChatGPT picks up on recent work flow to continue unfinished tasks, and reads recurring work patterns to proactively suggest skills or scheduled tasks.
The released image included a video thumbnail with the caption "ChatGPT Learns How You Work." It appears to have been shot at a desk with a laptop and mug, suggesting it's a feature introduction video. However, the brief announcement itself did not include specific screen paths or detailed setting options.
What "understanding work context" means
It means ChatGPT uses a user's recent work history as reference material. For example, if you stopped coding midway through a Codex session, reopening it would first identify where you left off. If a pattern is detected — such as writing the same weekly report or organizing data — ChatGPT is expected to proactively ask things like "Should I save this as a skill?" or "Should this run automatically at this time every week?"
How this differs from last week's sync announcement
As METAL LAB confirmed and reported on August 11, OpenAI had already rolled out a sync feature to the ChatGPT desktop app that lets users bring over projects, conversations, skills, and plugins between ChatGPT Work and Codex, with an option to enable automatic updates. This Computer History feature appears to be an extension of that. While sync was about "bringing over outputs created in another agent," Computer History is closer to "learning the context of what the user has actually been doing."
| Category | August 11 Announcement | August 13 Announcement |
|---|---|---|
| Feature name | Work Sync | Computer History |
| Target | ChatGPT Work, Codex | Codex, ChatGPT |
| Core content | Import projects, conversations, skills, plugins | Understand recent work context, suggest skills/scheduled tasks |
| Delivery method | Desktop app, opt-in auto-update in settings | Opt-in (detailed screens not disclosed) |
How to try it
The announcement did not specify an exact menu path. However, as with last week's sync feature, it's likely to be an opt-in toggle in the settings screen. Here's what's confirmed so far:
- Turn on the Computer History feature (opt-in) — if not enabled, everything works as before.
- When you reopen Codex or ChatGPT, you'll be presented with a state that references your recent work context so you can continue from there.
- If you repeat the same task, ChatGPT will offer to save it as a skill or register it as a scheduled task.
For example, someone who writes the same-format issue tracking document every morning might, after a few repetitions, be offered the option to register that task as a scheduled one. It also appears that closing and reopening a session mid-coding-task will let users pick up exactly where they stopped.
So what changes
Until now, every time you opened ChatGPT or Codex, you had to re-explain the context from scratch — this feature reduces that repetition. This could be especially meaningful for tools like Codex, where coding sessions run long and re-explaining "where I left off" was itself an inefficiency, potentially leading to real reductions in work time. That said, since ChatGPT is using computer usage records as training material, users should check for themselves how data is stored and deleted after opting in.
