Non-Meta-Problems
Figuring out what to work on is hard.
In hindsight, many of history's best opportunities seem random. You couldn't plan the discovery of penicillin, or predict the growth of the internet. And yet working in one field or another, one company or another, one team or another, would've made all the difference.
Over time, this will only become more pronounced. Tech increases leverage, which increases the rewards of choosing the right option (and the cost of taking the wrong ones). The ground is shifting underneath us, and the path forward has never been narrower.
How to deal with this uncertainty? A common response is to focus on meta-problems: the problem of what problem to work on.
In theory, solving these meta-problems lets us rule out unpromising opportunities, increasing our odds of doing great work asymptotically toward 100%. But in excess, this process turns into an endless, impossible search for certainty. A way of keeping the option of everything, without ever becoming anything.
Is there a balance? And if so, how do we strike it?
The line between problems and meta-problems can be blurry.
Everything we do informs everything we do next, so there isn't always a clear boundary between "working" and "figuring out what to work on".
This blurriness puts us at risk of doing fake work: pouring effort into things that may look useful, but have no real-world feedback loops.
So, what if we did the opposite of that?
If fake work has no feedback loops, then real work is only feedback loops. It keeps us in contact with nature, or human nature, or both, and lets reality correct our errors.
To the extent that work has these feedback loops, it's real. And to the extent that it's real, we know we're dealing with non-meta-problems.
But the question remains: what to work on?
Life is finite, and options are perishable. So we're forced to risk time and effort on projects, with no guaranteed payoff.
The decisive skill, then, isn't picking correctly, but iteration speed. The pace at which we can work, generate new data from contact with reality, and adapt accordingly.
Of course, it'd be nice if we could just pick the right direction from the outset. But things change too quickly for this to be possible — by the time we've identified the correct path, it's already shifted. There's some value in making a decent guess, but the faster the world moves, the faster these returns diminish.
So "the right direction" isn't a direction at all. Instead, it's a process.
I wrote this essay to figure out if it was worth writing more of them.
To keep iteration speed high, I'll only cover important, thorny problems that I can't otherwise figure out.
Beyond that, all that's left is work.
I have a good feeling. We'll see if I'm right.
September 2026