All models are imperfect, but some are useful

I recently heard the phrase "all models are wrong, but some are useful" at a lecture - it's referring to statistical models (and can be applied to life generally) but there's more to it that - it removes the illusion that there’s a perfect way to begin. It's just not possible.

Every model/approach to any problem you choose will have limitations. That’s unavoidable. The mistake is thinking you can design your way around that upfront. You can’t. The only real choice is which problems you want to have.

Many people fixate on the "some models are useful" part and keep searching for the right one - and to be clear, some are definitely more useful than others. Yet in doing so, they forget the first half entirely: all models are wrong. Waiting for one without flaws just means never starting.

The point of choosing a model isn’t to eliminate problems - it’s to surface them. Once you pick one (not just any model!) and put it into motion, you get feedback. You see what doesn't work. And then, to borrow Mark Manson’s idea, you can start exchanging your problems for better problems.

Your first model might be crude. It will be incomplete. But it exists, and that’s a best start you can expect. Because usefulness doesn’t come from perfection - it comes from iteration. You're replacing unknown problems with known ones, and trading theoretical risks for real answers you can actually work with.

So don't buy one more book on the subject - start with that imperfect model. Accept its flaws and use it long enough to understand its problems. Then iterate on it for better one with the new information you learned.