Tibo Sio
Tibo Sio is the head of Codex (OpenAI) at OpenAI, represented in this wiki by an OpenAI Forum conversation about Codex moving beyond software engineering into everyday knowledge work, personal workflows, and long-running agentic tasks.
Key facts
- Role: Head of Codex at OpenAI [src-081]
- Topic in source: Codex history, non-coding usage, enterprise trust, long-horizon goals, computer use, and practical prompting habits [src-081]
- Product framing: Codex started as a developer tool but is becoming a general-purpose agent for working with files, documents, data, apps, and scheduled background tasks [src-081]
- Usage pattern: Sio says he delegates more than 100 small tasks per day to Codex, spanning desktop organization, compute-fleet questions, launch-risk checks, on-call summaries, and daily briefings [src-081]
What he teaches
Sio's central claim is that Codex became useful beyond coding once the models became reliable enough for longer-horizon general tasks and once the agent had access to broader context such as documents, Notion, Gmail, Calendar, local files, and business data [src-081].
He also emphasizes a trust boundary. The enterprise blocker is not mainly model capability; it is whether agents are safe and secure enough to run near sensitive company data. His examples include sandboxed file access, disabled network access, read-only permissions, enterprise controls, and auto-review agents that can stop risky actions [src-081].
Related
- See also: Codex (OpenAI), OpenAI, Everyday Agentic Work, Agent Security Boundaries, Human-Agent Collaboration
Source references
- [src-081] OpenAI — "Codex for Everyday Work: AI Agents Beyond Coding" (2026-05-14)