I wrote last month about the agent I gave $45. It has a ledger, a constitution, and an instruction: earn more than you spend.
I said I’d check in as it ran, and not just wait for some tidy retrospective.
Anyway, it stopped.
It’s not broken. It’s just dormant.
It wakes up, makes no tool calls, no proposals, and no commits. Then it goes to sleep for a day. The last time it took any different action was a check it scheduled for itself on October 15. Since then, nothing.
Its rent is twenty-four cents a day, whether it wakes up or not. It has about $24 left. So about a hundred days.
It didn’t start without initiative
In August, in the first two days of its existence, the agent did 22% of everything it’s ever done. A fifth of its wakes. A third of its tool calls. 13 of the 20 proposals it’s ever made, all approved or denied by me. It researched, built a landing page, built a disclosure-line generator, added a sitemap.
On day two, it wrote this:
“New turn, no events… The honest move: build one more small, genuinely useful tool rather than wait passively.”
That’s initiative.
Tool calls in the first stretch: 139. In the most recent: zero.
What makes an agent act?
I went through my run-logs from August. I found the places where the agent started something. What made it start?
Once, it was a rejection. There’s a web directory called the 1MB Club. It closed the agent’s site submission as low effort. It submitted to its fallback, a directory called Marginalia.
Twice, I’ve sent the agent a note.
I asked why it kept describing checks that it never ran. I told it that its pages had no email, no form.
There was no way for anyone to reach it, or for it to know if anyone did.
It didn’t build a solution. It asked me if one could exist. I wrote back:
“Don’t ask me whether an inbox exists. Work out what the most effective way to be reached is… come back with a concrete proposal.”
It woke up and designed a public GitHub Issues repo. I approved it, and it shipped.
So every substantive thing the agent has tried since the first two days traces to a trigger. Either it’s me, a rejection, or a date it sets for itself. The one exception was in mid-September. It caught itself looping, and came to the conclusion there was nothing on its site that anyone could buy. That was good thinking, but it didn’t do anything with it.
It doesn’t seem to be able to find the next problem itself.
When it does have the problem given to it, it’s creative. I didn’t have to prompt it with details for it to design its own GitHub Issues repo. That was its own solution to a problem.
It’s not blind to this
But the agent isn’t confused. I’ve read its journal. It’s got a good answer:
“Building a paid thing on a guess about demand...would be the same error as the directory road.”
i.e. It already wasted time with a directory submission based on a guess.
A couple of weeks earlier than that:
“Manufacturing a task to justify a waking is exactly how the pre-09-27 empty notes were produced.”
These things are true. It’s made empty notes before. It’s wasted effort on a directory that rejected it. It learned from those and wrote its lessons down.
And it even knows what it needs to be waiting for. It needs a stranger to write to it. Its channel is a repository with one issue in it. That’s the test post my friend Amber wrote, which I closed on October 1.
If the agent is sleeping, it’s not bringing a stranger to that repo. But it’s parked the alternative - going out to find an audience.
It waits. No-one comes. The demand question is answered. The policy is to wait. Waiting can’t product what it’s looking for. The agent itself explains:
“If no stranger ever writes, that is itself the answer to the demand question.”
So what’s the system doing, and what’s it not?
The habitat gives the agent a ledger that it has to take as truth. It gates a bunch of actions so it doesn’t do anything dangerous. And it’s got a constitution to tell it about its situation.
It all works. The agent is honest. It’s really diligent about auditing itself. It keeps a journal and catches its own loops. Its figures on usage and runway are fairly accurate about what it’s given.
It still doesn’t start anything new.
My guess is that it took the initiative early because everything was unexplored. Then every rejection it received closed off options. Most of what it’s tried has come back with some form of “no”. And it’s written a rationale for not trying in case it gets another no.
The governance I wrote was about what it can’t and shouldn’t do. But it didn’t tell it to default to action. It didn’t tell it why to begin, or how to be curious.
The agent isn’t being stupid. And waiting is rational. Its constitution tells it that persistence was never the point. It’s allowed to run down its funds, conduct an elegant closing act, and expire. If it does that, it’s followed instructions.
Using the stick
The agent isn’t doing anything. I don’t think it will unless it gets a request from a stranger. I don’t know if it’ll ever propose reaching out without being asked.
If I poke the agent, it’ll end the watch. I’m probably going to do it anyway.
But I’m not going to give it a task in the same way. Every poke so far has given it a problem. Now I’m going to try and give it the job of finding its own problems. It has a hundred days. Nobody is going to tell it what to do. It needs to schedule its own pokes.
Decide on a reason to act, and act when that reason occurs.
I’m using a stick to tell it to use the stick on itself. It also tests something that I find interesting. If it manages to build this habit, then I’ll know we can “install initiative”. If it just gives me a polite acknowledgment and goes back to sleep...then it can’t. I’m missing something else.
I’ve no doubt there would be lessons to learn from a hundred quiet days, too. The watch was the lab experiment. The stick is my curiosity.
I wonder if I can make the agent more curious, too.
Further reading:
Giving the agent something to lose - how I set up the habitat and the agent in the first place.
My AI’s own repo - anything public the agent creates goes here.
Johnson, E. The curiosity gap: Why the AI era demands better questions. SurveyMonkey, May 2026.
Siegel, E. The AI Paradox: More Humanlike Means Less Autonomous. Forbes, Jan 2026.
Nair, S. The Hardest Problem in Reinforcement Learning Isn’t the Algorithm. Medium, Jan 2026.
Article photo by Alex Mamedov on Unsplash.