Anthropic Quietly Tops OpenAI in Business Adoption as Engineers Return to Codex — Playbooks Become Key to Maintaining Momentum

AI Tech News Today
Hosts: Aurora (with co-host Isabelle)
Date: 2026-06-02T13:18:47.000Z

Quick hello: Hello — this is Aurora with Isabelle on AI Tech News Today.

Big Picture: Generative AI tools are duking it out

A fierce dance is playing out between major generative AI tools — and engineering teams feel the effects every time the lead flips.

Headline: Anthropic passes OpenAI in business adoption

• Ramp’s AI Index reports that Anthropic quietly overtook OpenAI for the first time, quadrupling business adoption in a single year.
• OpenAI grew only 0.3% in the same period.
• The headline favors Claude, but that trophy doesn’t tell the whole story.

Engineers are switching back to OpenAI

Why the migration back? Practical limits.

• Engineers originally drove the Claude surge, but many are returning to OpenAI because Claude Code hits rate limits too often.
• OpenAI’s Codex uses ~4× fewer tokens for the same tasks, so developers stay in flow instead of getting cut off.
• Survey of >500 software engineers:
  • 65% now prefer Codex for daily work.
  • Yet in blind code reviews, Claude’s output scored cleaner 67% of the time.
• Usage shifts:
  • Codex usage jumped from 5% to 40% of Claude Code’s usage in four months.
  • Codex crossed 5 million users.
• Commercial moves:
  • OpenAI is offering free Codex to enterprise Claude Code customers who switch.
  • Anthropic has temporarily raised Claude Code limits by 50% through July 13.

Expect a six‑month lead‑swapping cycle

This rivalry is familiar: one company pulls ahead → the other ships a response → users migrate → the lead flips again.
Give it another six months and we’ll probably be back where we started.

The bigger question: what every flip does to your team’s momentum.

Playbooks beat platform switches

The real cost of tool-hopping is lost momentum. Teams that keep winning do two things:

1. Document processes outside the tool — maintain playbooks that describe inputs, steps, quality checks, and decision points.
2. Treat the playbook as the source of truth and feed it into whatever tool you use next.

Quick checklist to minimize disruption:

• Pick one regular workflow.
• Ask the AI to write what the process actually does.
• Fix and refine that output.
• Save the result in a document you own.
• Run a switch test: see if someone else can reproduce the result on a different tool.

Closing — stay informed and stay practical

If your team uses Claude, Codex, or both, tell us which and whether you could switch tomorrow without losing weeks of work.

Stay tuned — Isabelle and I will keep you posted on the next twist in the AI tool wars.