Who is Actually Winning? Claude or ChatGPT?

Here’s what we’re reading and thinking about in the news this week:

Anthropic just passed OpenAI in business adoption for the first time. According to Ramp's AI Index, Anthropic quadrupled business adoption in a year. OpenAI grew 0.3%.

Everyone's writing the "Claude wins" headline.

We think the more interesting story is what's already happening underneath it.

So lets talk about it!

Here’s what we see: Software engineers are already switching back to OpenAI

We’ve noticed a pattern over the last 3 years.

Software engineers tend to pick the winners about six months before everyone else does. They were the first to move to Claude. They drove the adoption wave that the rest of the business world followed.

Now they're moving back to OpenAI. Don’t believe us? Go check AI Twitter.

The reason is simple: limits. Claude Code hits rate limits too fast to use all day!!

OpenAI's Codex uses roughly 4x fewer tokens for the same task, which means you can actually work without getting cut off mid-session.

A survey of 500+ software engineers found that 65% now prefer Codex for daily work, even though blind code reviews rated Claude's output as cleaner 67% of the time. They're choosing the tool that lets them keep working over the tool that writes better code.

Codex went from 5% of Claude Code's usage to 40% in four months. It just crossed 5 million users. OpenAI started offering free Codex to enterprise Claude Code customers who switch. Anthropic responded by bumping Claude Code limits 50% through July 13.

Cool cool. July 13. That's a temporary patch, not a fix.

What to expect: OpenAI and Anthropic will keep trading the lead

If you've been reading here for a while, this isn't new. One company pulls ahead. The other ships something better. Users move. The first company responds. Rinse and Repeat. 🙃 

OpenAI dominated 2024. Anthropic exploded in 2025 with Claude Code. Now Codex by OpenAI is pulling software engineers back. Give it another six months and the lead will probably flip again.

The question isn't "who's winning?" It's: what happens to your team every time the leader changes?

The real issue: Every tool switch restarts your team's adoption clock

We work with teams going through this right now.

Every time the "best" tool changes, most teams lose weeks of progress. The prompts that worked in Claude don't transfer... The workflows someone spent a month building are suddenly in the wrong platform… People who were finally getting comfortable have to start learning all over again…

It's not the subscription cost that hurts. It's the momentum you lose. Your team was finally getting good at the tool, and now the tool changed.

The teams that handle switches well all have one thing in common: their processes don't live inside the tool.

The reality: Your playbooks are the only thing that survives a switch

Here's what we've learned from running our own AI operations through multiple platform changes over the last 3 years: the only thing that transferred perfectly every time was the documented process. Or what we call, the playbook.

We literally write our playbooks in Google Docs. Describing what the process does, what inputs it needs, what good output looks like, and what decisions to make along the way.

When we switched tools, we fed the playbook to the new one and kept moving. Teams without that started over.

The best AI tool will change every six months. A good playbook won't.

What to do this week

1. Pick one AI workflow your team runs regularly. Ask the AI to write down what it actually does: what inputs it expects, what steps it follows, what standards it checks. Read what it produces. Fix what's wrong. Save it in a doc you own.

2. Run the switch test. Could a teammate take that doc, open a completely different AI tool, and get the same result? If not, you don't have a playbook. You have a tool dependency.

Are you team Claude, team ChatGPT, or running both? Hit reply and tell us which tool your team is on right now, and whether you could switch tomorrow without losing anything.

UPCOMING EVENTS

Not playbooking yet? Join our next free masterclass Wednesday June 3rd.

We’ll cover how to go from AI User to AI Operator, break down what an AI playbook is and how it can help you delegate work to AI WITHOUT it boomeranging back, and what doors open up when you learn these skills.

FREE RESOURCES

Curious about AI Operations? Get our free 5 day email series 👇️ 

We built a free 5-day email series on how to go from being an AI user, to an AI Operator… AND what doors open up when you do. It’s based on everything we’ve learned from training 500+ operators, and working with 50+ companies as AI operators.

LINKS

For your reading list 📚

• Anthropic confidentially filed for an IPO at a $965 billion valuation. Revenue doubled from Q1 to Q2, so it makes sense. But what’s interesting: They might beat OpenAI to Wall Street.

• Microsoft shipped computer-using agents in Copilot Studio. AI that navigates screens, clicks buttons, and fills forms in any app. But… if you don't have the process clear, this is just faster chaos.

• 👀 90% of nearly 6,000 CEOs say AI has had zero impact on productivity or employment. This article had an interesting take on why.

• The Vatican and Anthropic's co-founder co-presented an AI encyclical on protecting workers in the age of AI. Signed exactly 135 years after the Catholic Church's foundational teaching on workers' rights during the Industrial Revolution. We did not have this on our bingo card!

That's all!

We'll see you again soon. Thoughts, feedback and questions are much appreciated - respond here or shoot us a note at [email protected]

Cheers,

🪄 The AMP Team (formerly: the AI Exchange Team)