Fail Forever.

15 Sep, 2026

“Two obsessions are one too many.”

~James Watson

There is a boring interpretation of this, which is just time allocation. You have a finite amount of time,

so every hour spent on one thing is an hour you cannot spend on another. This is true, but also not very interesting.

I think obsession is stranger than that. It is not merely spending more time on something. It is allowing the thing to occupy enough of your mind that eventually you stop having to deliberately think about it. You read something completely unrelated and somehow it connects back to the problem. You are walking somewhere and something clicks. You wake up with an answer to a question you were stuck on yesterday. At some point the boundary between working on the thing and not working on the thing becomes unclear.

This is probably not an especially "balanced" way to live. It also seems to work.

There is some function

relating time invested to performance. The usual productivity-brain interpretation is roughly linear:

Put in twice the hours, get twice the output. Allocate half your time to two things, become half as good at both. Very civilized. I don't think sufficiently difficult things work like this.

Suppose, just as a toy model, that over the relevant part of the curve,

Then dividing your attention equally between two obsessions gives

Since

we get

and therefore

Obviously human performance does not literally obey this equation. The point is simpler: depth compounds.

The thousandth hour contains things the hundredth hour could not contain because it sits on top of the previous nine hundred. You have more context loaded. More failed approaches. More weird little facts that didn't matter when you learned them and suddenly matter enormously. Eventually you begin seeing structure that is effectively invisible to someone arriving fresh.

Two half-obsessions therefore may not add up to one obsession. You lose something in the split.

So Watson's prescription seems simple enough: pick one thing and become unreasonably obsessed with it. Unfortunately there is a fairly serious bug in this strategy;

You don't know if you picked the right thing.

“If you want to change the game, if you want to change mankind, if you want to do something seminal, you have to be willing to fail forever.”

~Peter Attia

I keep coming back to Attia's words. Forever is doing an incredible amount of work in that sentence. It does not mean work hard until you succeed. It does not mean grind for five years and eventually the universe, impressed by your discipline, deposits the appropriate reward into your account.

It means you might not succeed. Ever.

If you are trying to do something genuinely new, the feedback loop can be horrifyingly long. The experiment might require ten years before producing enough information to tell you whether the original hypothesis was stupid.

And the experiment is you.

You can dress this up as expected value:

But the notation hides the only variable that actually matters.

For a sufficiently serious obsession,

You don't get the principal back.

This creates a very strange optimization problem because the obvious response to uncertainty is diversification. If you don't know which bet will work, make several bets. Except Watson's entire argument is that this may destroy the thing you are trying to optimize.

If

then allocating time to $B$ and $C$ protects you against being wrong about $A$. But that same protection may prevent you from ever discovering what happens when

So the wager has an unpleasant structure. You have to concentrate before you possess enough information to know whether concentration is rational.

That is obsession.

Maybe you spend twenty years on the right problem and change the field. Maybe you spend twenty years becoming the world's greatest expert on something that did not matter. Maybe the problem was impossible. Maybe someone solves it six months after you die. Maybe you were one lemma away. Maybe you were nowhere close and simply became extremely sophisticated at explaining why everyone else failed to understand your genius. The worst part is that from somewhere around year eight, these lives can look remarkably similar.

From the inside,

Both wake up thinking about the problem. Both continue after reasonable people would have stopped. Both have increasingly elaborate explanations for why the thing matters. Both can point to small pieces of evidence suggesting they are getting closer.

Only one eventually gets the luxury of having the obsession retroactively renamed vision.

This is why I don't think obsession is inherently admirable. Obsession has no truth condition. You can be obsessively right and obsessively wrong. The intensity of your belief contains almost no information about its correctness.

If anything, obsession is an amplifier:

while

Which makes the usual advice to “find your obsession” feel backwards. The scarce thing is not obsession. There are plenty of obsessive people. Give a sufficiently strange person a sufficiently interesting problem and obsession is almost the default outcome.

The scarce thing is judgment.

What deserves twenty years? What problem is important enough that failing on it would still have been a reasonable use of the time? What are you willing to continue doing when the external reward function remains approximately

for years?

And, more importantly, what are you willing to fail at forever? Watson says two obsessions are one too many. Attia adds the much more frightening condition: the one you choose may consume an enormous fraction of your life and give you nothing back.

There is no theorem telling you which problem deserves the bet.

You have to pick anyway.