Last week Anthropic published an essay called “When AI Builds Itself.” Three million people read it within a day. I read it twice.

The argument is simple. AI research is the first field where AI is being applied to itself at a superhuman level. Anthropic’s engineers ship eight times more code per day than they did in 2024. Claude wrote more than 80% of the code Anthropic merged into its own codebase last month. The length of tasks AI can do reliably has gone from doubling every seven months to doubling every four. Two years ago, Claude could handle a four-minute software task. Today it handles a twelve-hour one. If the trend holds, tasks that take a person weeks will be in range by 2027.

That is the research lab. But the line that should stop you, if you are a senior executive with 25 years inside finance, consulting, pharma, tech, or law, planning your exit in the next three to six months, is this. Whatever happens inside AI research over the next eighteen months is what happens to your field one to three years behind it.

That is the canary.

Here is the tension nobody is naming clearly enough. Most of the people I speak to are watching this from inside a thousand-person organisation. They feel the heat. They see the AI memos. They sit through the strategy off-sites. And the conclusion they draw is something like, I should learn more about AI, push my team to adopt it, maybe sign up for a course.

That is the wrong size of response.

Anthropic’s essay lays out three scenarios. In the one they call most likely, AI capabilities keep compounding and a hundred-person company starts doing the work of ten thousand. Each person sits on top of a pyramid of AI agents. This is not a hypothetical for 2030. The conditions for it exist today.

If you sit inside a legacy thousand-person business, your position is the position of the business itself. Both of you are about to be outcompeted by something smaller, faster, and built AI-native. Your raise, your promotion, your bonus, your equity, are all riding on a structure that is being quietly priced for replacement.

The reframe I want you to sit with is this. The same force dissolving your seat is the force that finally makes the independent path real.

For twenty years, the reason senior people stayed corporate was the team and the engine behind them. The bank, the consultancy, the pharma group, gave you analysts, a brand, a sales pipeline, a back office, an IT department. Stepping out meant trading all of that for being a one-person shop with a website.

That maths changed sometime in the last fourteen months. The person leaving today opens her practice with a team of AI agents already handling research, content, outreach, qualification, proposal drafting, scheduling, follow-up, invoicing, and reporting. She is the hundred-person company Anthropic is describing. The expertise is the moat. The AI agents are the engine.

I have watched this play out with the people I work with. One of my clients ran her first five-figure month before her notice period was up. Another did the same inside ninety days of leaving her corporate role. They moved through more weekly output than they ever managed inside their old firms, with AI agents doing the volume work that would have required hiring a small team in 2022.

If this is hitting close to home, here is what I would actually do this week.

Take one hour and read the Anthropic piece end to end. The three scenarios at the bottom are the part that matters for you personally.

Then sit with one prompt. Imagine a hundred-person AI-native version of your current company exists today. Find where you sit inside it. In most cases the picture is one of three things. The role does not exist. The role exists at one tenth the headcount. Or the role looks exactly like what an independent expert with AI agents already delivers to clients.

If you land on the second or third version, the question is no longer whether to leave. The question is when, and what the first ninety days outside look like. That is a planning problem.

The window for this move is wider than it looks from inside the corner office and narrower than it looks from outside it. Eighteen months ago, building an AI-native independent practice was hard. Eighteen months from now, the field will be crowded with people who saw what you are seeing today and moved first.

The canary is singing pretty loudly. The question is what you do about it

Warmly,

Salama

If you are in the three-to-six-month planning window for your move and want to look at your specific situation with someone who has guided 100+ senior leaders through this exact transition, I set aside time each week for fifteen-minute strategy sessions. No pitch. A clear first step. Apply here.