In employment, the constraint is external. Your manager tells you what matters. The customer tells you what matters. The quarterly goal tells you what matters. You disagree sometimes, you optimize locally when you can, but fundamentally: someone else has already decided.

I spent years working within those constraints. They were useful constraints. They shaped the work. They meant I never had to ask “is this worth doing?” because that question was already answered. I just had to answer “can I do this well?”

Since 2023, I haven't had that. There is no manager. There is no customer. There is no quarterly goal. There is only pro bono independent research, funded by the thinnest possible margin: a small amount of state support and the principle that the work matters enough to keep doing it anyway.

The freedom is real. I can pick the questions I actually care about. I can spend three months on something that might produce nothing. I can say no to work I don't believe in. I can pursue lines of research that nobody is paying for because I think they're important.

The terror is equally real.

When nobody is paying you, when there's no team, when there's no deadline except the one you set yourself—how do you know you're right? How do you know you're not just researching whatever is most interesting to you rather than whatever matters? How do you distinguish between “I think this question is important” and “I like the dopamine hit of following this rabbit hole?”

In employment, that problem was solved for you. The customer is always right, even when they're not. But in autonomy, you have to be right and be able to admit when you're wrong. You have to have something other than “I wanted to” as your justification.

The work itself has changed as a result. It's deeper, I think. I write things now that I wouldn't have been allowed to write in employment, not because they're secret, but because they take longer and produce less obvious value. I've disclosed security findings that didn't have easy metrics attached. I've done methodology work that nobody asked for because I thought the field needed it.

But I've also spent time on things that didn't work. I've chased hypotheses that went nowhere. I've written papers I never published. I've built tools nobody uses. I've made mistakes in public because I was trying something unfamiliar and got it wrong.

In employment, you hide those. Your manager doesn't see the failed attempts; she sees the shipped feature. The customer doesn't see the wrong architectural decision that you pivoted away from; she sees the product. The value is polished before it's presented.

In autonomy, the failure is often visible. Sometimes that's the point—transparency about methodology means showing the failed approaches. Sometimes it's just... failure. Overhead that eats time and resources and doesn't yield anything.

The honest part is that money is tight. I'm not wealthy. I have a small safety margin, and that margin gets thinner when the research takes longer than planned. There's a constant calculation underneath the intellectual work: can I afford to keep pursuing this? How long until I need to find paying work? What happens if the funding landscape shifts?

The other honest part is that the focus is harder without the external constraint. It's easier to stay off a rabbit hole when your manager is waiting for a status report. It's easier to say “this is good enough” when someone else is paying the bill for perfection.

But the quality of the work is higher. The things I publish now are things I actually believe in, not things I was assigned to produce. The findings I disclose are findings I thought were important, not findings the contract required. The papers I write are the papers I think need to exist.

That matters. That might be worth the tighter budget and the harder focus and the constant question mark hanging over how long this can last.

The interesting question isn't whether autonomous research is better than employment. It's different. It trades security for integrity, and money for agency. It trades the clarity of external goals for the terrifying freedom of setting your own. It assumes you have the discipline to choose right when nobody would know if you chose wrong.

The sustainability question is the one I can't answer yet. I don't know if this is sustainable. I don't know if I can maintain this quality of work, this level of integrity, indefinitely on a diminishing budget and an uncertain funding path. I don't know if the contribution is large enough to justify the cost.

What I do know is that the work is better. What I do know is that I'm braver about the questions I ask and honester about the answers. What I do know is that the research is mine, and that changes everything about it.

Whether that's enough—whether it's sustainable, whether it matters, whether the work I do in this mode will outlast the money that funds it—that's the unanswered question. That's the sabbatical, really. Not a break from work. An experiment in what happens when you remove the external constraint and ask someone to decide what matters.

I'm still finding out.