the gap between the outcomes people get who prompt well vs the ones who don't is growing larger.
the top 1% thinkers, builders, and engineers have a larger lever through ai than the average person. purely because they use ai in different ways.
a junior asks "build an auth system." a senior asks "add session-based auth with jwt, rate limits, token rotation..." — the difference in output quality is enormous, and it compounds.
the consequence of this is that we will have value generation and capture that grows even more asymmetrically between people. this has always been the case, but ai accelerates this shift in the distribution.
the founder who understands their market, customers, and dynamics more deeply will kill the founder who lives in the illusion of "understanding."
so a natural question worth asking is: what makes someone good at prompting?
to me, prompting is knowledge.
the reason my prompts for building things are better than some average joe's is because i can describe with clarity and depth what i want. which is a product of my knowledge. this includes judgment and taste.
how do you get knowledge? by doing things and learning. experience, simply put. not passively observing what the model is spitting out.
the issue is that many tend to outsource their thinking to the model. and get lazy in acquiring deeper, more specific knowledge because they take the route of least resistance.
it is not obvious when you do it, but this approach leads you to acquire "pseudo knowledge" — which looks fine at first glance, but is just slop when you inspect it more deeply.
the way i think about ai these days is that it's just leverage for my mind. and if my mind has bad ideas, the outcomes will be bad.
you are in control of the depth and rigor you apply when attaining knowledge. thus, you are in control of the curve you are on in the distribution.
which is good news. but requires you to use your brain and actually work.
this might change in a few years (or months), but right now prompting is knowledge. and knowledge is what allows you to get abnormal results. acquire deep knowledge.