Writing
Knowing vs Figuring Out: Why Humans Still Matter in the Age of AI
AI is getting very good at knowing. Give a model a well-posed question, the right context, and a clear target, and it will usually produce a better answer than most people could. I work with these models every day, and they know more than I do about almost everything. Yet my job has not gotten easier, because the job was never knowing. The job is figuring out: deciding which question matters and what would count as a good answer.
The distinction sounds small. It isn't. Knowing operates inside a frame somebody already built. Figuring out builds the frame — it notices the hidden constraint, the unstated incentive, the metric that's quietly measuring the wrong thing, the stakeholder nobody invited, the reason the obvious solution won't survive contact with production.
I see this constantly in moderation work. A model can classify content remarkably well once you've defined what counts as a violation. But deciding where that line should go — what the policy is, which edge cases break it, how it interacts with the incentives of the people creating the content — no amount of model output settles that. The classification is knowing. The policy is figuring out.
Why figuring out stays scarce
Knowing is retrieval and synthesis, and models have made both nearly free. Figuring out is framing, compression, causal reasoning, and — the part people skip — deciding what evidence would change your mind. A model will generate candidate frames for you all day. It will not tell you which frame your particular messy organization, with its particular incentives and history, actually needs. Choosing the ontology is still on you.
So the human advantage shifts toward problem formation. When someone hands you a question, the useful move is to ask what decision the answer is supposed to inform — half the time the real question is hiding behind the asked one. Separate what you know from what you're assuming from what you'd merely prefer to be true. Then let the model search the space fast. That's the division of labor that works: you choose the frame, it explores the territory.
One more reason to take this seriously: the reasoning behind your decisions is becoming infrastructure. Future agents will need the why as much as the what, and the why is exactly the part that never makes it into the ticket. Writing it down used to be good hygiene. Now it's leverage.
As knowing gets cheaper, figuring out becomes the scarce work. The people who can frame a problem will have more leverage than they've ever had. The people who were valuable mainly for knowing things are about to be in a tough spot.