OpenAI’s Codex shift gets practical

Codex is becoming the work layer

Codex left engineering.

OpenAI says Codex now accounts for 99.8% of weekly output tokens inside the company, and every department uses it as its primary AI tool for work. The paper says individual users are also giving Codex longer jobs, with 70.2% making at least one request estimated above one hour by May 2026.

This stands out because adoption moved fastest once the tool stopped feeling like a coding product and started handling messy office tasks. I buy that signal, especially because Legal, Finance, and Recruiting crossed into majority Codex use around April 2026.

The desk-level change is easy to picture: a recruiter at 4 p.m. asking an agent to clean data or build a small tool.

The next question is whether companies redesign jobs around this, or just pile agent work on top of the old work.

That tension feels real.

Meta staffs up for agent risk

Meta is staffing security.

After Anthropic’s Fable 5 and Mythos scrutiny, Meta is pulling three Virtue AI founders into Superintelligence Labs. Bo Li, Dawn Song, and Sanmi Koyejo are joining alongside other team members from a startup focused on red teaming, guardrails, and AI governance.

That timing matters. Meta is building agents for billions of users, so security can’t sit off to the side as a review step after launch.

I read this as a practical admission that agent safety is becoming an execution problem, not only a research problem. The work sounds like people at desks testing failure cases, reading logs, and tightening controls before products reach users.

It also shows how AI labs are recruiting: they want compact teams with specific skills, even when they do not buy the whole company.

The open question is whether this talent lands early enough to shape the products, or only clean up after them.

AI’s price tag is getting harder to ignore

The AI data-center boom is starting to show up in prices people actually pay.

WSJ says demand for memory chips is lifting costs, with pressure spreading toward smartphones, computers, and electricity. The piece frames this as a possible third wave of inflation after trade tensions eased and gas prices started falling.

That matters because AI spending is no longer just a tech-industry story. It is moving into the normal price system, where a $200 increase on Apple devices or a higher power bill can change household behavior.

I think the key question is timing. If AI productivity gains arrive slowly, consumers may feel the costs long before they see the benefits.

The quiet detail here is physical: more servers, more chips, more power lines, more strain at the outlet.

That makes this less about AI promise and more about who pays first.

Prohuman team

Free Guides

Explore our free guides and products to get into AI and master it.

All of them are free to access and would stay free for you.

Feeling generous?

You know someone who loves breakthroughs as much as you do.

Share The Prohuman it’s how smart people stay one update ahead.