Three people lead 100 AI programmers and burn through $1.3 million in a month! OpenAI: I'll cover the cost.

Peter Steinberger

Editor: Solomon

[New Intelligence Yuan Introduction] Three people, 100 AI agents, burning through $1.3 million in a month - the father of OpenClaw has turned software development into an "AI assembly line", and OpenAI is footing the bill.

While others show off their pay stubs, he shows off his bill - $1.3 million a month!

That's nearly 9 million RMB per month. It really stunned netizens.

Peter Steinberger, the father of OpenClaw, casually posted a screenshot on X.

Peter Steinberger

But the numbers on the screenshot are really eye - opening:

Cost in 30 days: $1,305,088.81. Consumed 603 billion tokens. Initiated 7.6 million requests.

You read that right, it's $1.3 million. And it's not the quarterly AI budget of some big company - it's the usage of a three - person team in a month.

Even more astonishing: OpenAI is reimbursing this expense.

The comment section went crazy instantly.

Some were amazed, some were skeptical, and some took out their calculators to figure out "how many programmers this is equivalent to".

Steinberger himself calmly responded: "After turning off the fast mode, my cost is lower than that of an engineer, and it really helps a lot more."

In other words - it's really cost - effective!

Some netizens were even more amazed at the engineer with a monthly salary of $400,000 - "The job market in San Francisco is just too crazy."

Netizen comments

Netizen comments

Some netizens were curious about where all those tokens were spent.

Peter replied that most of them were used in the development of OpenClaw.

Netizen comments

The Cloud Programmer Legion

The most incredible thing about this is that Pete's small team only has three people.

They have about 100 Codex instances running in the cloud for a long time, doing the dirtiest, most tiring, and most maddening jobs in software engineering -

Reviewing PRs, finding security vulnerabilities, deduplicating issues, fixing bugs, monitoring benchmarks, posting regression findings on Discord, and even opening PRs right after a meeting.

So, AI is not just "helping you write code", but has penetrated into every aspect of software collaboration.

This is really scary.

Because the really expensive parts of software development are communication, understanding, context switching, review, regression, repair, waiting, and repetitive work.

In the past, a team spent a lot of time every day on things that don't seem like "creation", but without which the project would fall apart.

Now, Peter has handed all these tasks to a group of AI agents at once.

This is AI starting to maintain the nervous system of an organization for you.

Schematic diagram

There is an important detail in this screenshot: it's not the OpenAI backend, but CodexBar developed by Peter.

CodexBar is a macOS menu bar tool used to track the usage windows, credits, costs, and reset times of various AI programming tools.

It supports a bunch of services such as Codex, Claude, Cursor, Gemini, and Copilot.

What used to be in a programmer's menu bar? CPU, memory, battery, and network speed.

Now there's something new: tokens. Tokens are becoming a new kind of "means of production".

CodexBar

A Few Final Words

$1.3 million a month, three people, 100 AI agents.

Think about these numbers carefully - three real people, leading 100 digital employees that don't eat, sleep, or ask for a raise, are doing the work of an entire engineering team.

Some people are excited after reading this: AI is finally not just a chatty vase! Others feel a chill down their spines: Wait, what will we coders do in the future?

But to be honest, what really keeps me up at night is Steinberger's casual remark: "I'm exploring what software development would be like if token cost weren't an issue."

Peter Steinberger

Folks, he said "if".

But the problem is, this "if" is becoming a "when" at a visible speed.

The work that costs $1.3 million today will cost $130,000 after one round of price cuts for the model. After another round, it'll be $13,000.

By that day, having 100 AI agents working simultaneously won't be an exclusive game for Silicon Valley tycoons, but a basic operation for any three - person startup team.

Three young people in a garage, with 100 tireless AI programmers in hand - just imagining this scene is mind - boggling.

Peter Steinberger has revealed the cards.

What's written on the cards is: The future is knocking on the door, and it's not going to wait for you to get ready.

References:

https://the-decoder.com/for-1-3-million-a-month-openclaw-founder-peter-steinberger-runs-100-ai-agents-that-code-review-prs-and-find-bugs/

https://x.com/steipete/status/2055346265869721905

https://developers.openai.com/codex/speed

This article is from the WeChat official account “New Intelligence Yuan”. Author: New Intelligence Yuan. Republished by 36Kr with authorization.