Only 1 in 1,600 People Use Codex. Here’s How to Catch Up.
Nate B Jones frames Codex as more than a coding tool: it is a way to delegate full computer-based work loops to agents.
Nate B Jones presents Codex as a shift in how people use computers. The point is not just that the tool can write code, but that it can work across files, folders, browser sessions, documents and websites until there is a concrete result to inspect.
From answers to assignments
His rising token usage is framed as evidence of a larger change: the unit of work has expanded. Instead of asking for a draft or a summary, he gives Codex jobs such as finding sources, reading folders, rendering a document, checking that it opens and continuing until the goal is met.
Persistent threads and focused sub-agents
Jones argues against treating every AI interaction as a disconnected chat. A central “chief of staff” thread can hold the project goal, current artifacts, folders and standards, while narrower sub-agents handle discovery, source checks, output inspection or noisy-folder summaries.
A practical starting point
The recommended entry point is one valuable, annoying loop: turn a transcript into a brief, organize a source folder, prepare the day from calendar and messages, or build a personal dashboard. The setup needs five things: a goal, sources, a standard, permission boundaries and proof that the job is done.
Power requires boundaries
Because Codex can operate through real computer contexts, Jones emphasizes discipline: do not paste secrets into prompts, avoid unnecessary write access, be careful with publishing or deletion, and require receipts such as files, logs, tests, renders and command output.
The broader signal
The video frames Codex as early training for a new kind of computer literacy: delegating work to agents that can use the machine, then learning how to inspect and govern the work they return.
Source
- Chaîne: AI News & Strategy Daily | Nate B Jones
- Vidéo source: https://www.youtube.com/watch?v=xqGCbEDbny8