One Time David Sinclair is a Harvard biologist who has spent his career studying how the body ages. On a recent podcast, he said his lab did 160 years of research in two months.
Not because the AI got smarter. Because he could run a thousand copies of it at once, working around the clock, one AI brain teaching the others whatever it learned.
His lab is capped at twenty human researchers. He cannot hire past that, so he stopped trying and replicated the work instead. That is only possible with what the current tools can actually do.
Every small business hits the same ceiling
Yours is not twenty researchers. Yours is you, plus however many people you can afford, plus however many hours are left after the part of the job that only you can do.
If you are a real estate agent, you know this one. You have written a CMA from scratch for every listing. You have answered the same buyer questions about financing and timelines more times than you can count. You have followed up with every open house sign-in by hand.
You do it again, every time, because you thought nobody else could.
The word doing the damage in that sentence is nobody. It is true and it is also irrelevant, because the question was never who else could do it.
What if you copied your brain instead
That is what an AI agent is. Not a smarter you.
A copy of your best hour, running while you do the part of the job only you can do.
The CMA. The follow-up. The listing workflow. Not improvised each time by somebody guessing at your standards, but run the way you decided it should be run, on the day you were sharp enough to decide it.
That is the difference between hiring and copying. A hire learns an approximation of your judgment over six weeks and then leaves with it. A copy starts from your judgment and never drifts from it.
The objection, and why it is the wrong one
The usual pushback is that this is a bubble, that the tools are overhyped, that it is a lot of noise about something that mostly writes mediocre blog posts.
Some of that is fair about the marketing. None of it is an argument about the ceiling.
The ceiling is real whether or not any of this gets better than it is today. You have a fixed number of hours and a fixed number of people, and there is a category of work that eats both without needing either. That was true before any of these tools existed. It stays true if they stop improving tomorrow.
The only question that matters is whether you can hand a repeating job to something that will do it the way you would have done it. Today the answer is yes, for a real and growing list of jobs, and the list you personally can hand over is longer than you think.
Start with one copy
Not a thousand. One.
Take the single thing you have done from scratch more times than you can count. Write down how you do it, once, properly. Hand it over. See whether the output is the one you would have written.
When it is, you have not saved an hour. You have removed a ceiling.