solo builders already have the skill corp PMs are panicking about
i've been watching PM friends panic about claude design and it reminded me of something solo builders already know that corp PMs are about to have to learn.
you're the one who decides which of the ten AI-generated mockups ships. the AI doesn't tell you. figma doesn't tell you. claude design doesn't tell you. you pick.
that pick is a taste call. it's based on every bad version of the product you shipped before, every customer call that went sideways, every conversion drop you had to explain to nobody but yourself.
solo builders have been doing this forever. corp PMs used to be able to hide taste inside a committee. the committee took the blame for a bad call because "we agreed on it." now the committee is a subscription and the call is yours.
i think the interesting thing for solo builders this year is that your taste-making muscle is suddenly the most valuable muscle in the market. the people you're competing with at bigger companies are figuring out they never built it.
not a humble brag - honestly half the time my taste is wrong. but the reps compound and i'd rather be wrong and learning than right and committee-approved.
anyone else feeling this shift? or am i just reading my own weekend into the zeitgeist.
This is a great way to frame it — the “taste” layer is becoming the real leverage.
AI compresses execution, but it doesn’t compress judgment. If anything, it amplifies the cost of bad taste because you can ship faster.
Feels like the gap now isn’t skill vs no skill — it’s clarity vs noise.
Curious how you think about developing that taste intentionally, vs just compounding reps over time?
reps give you the material, friction refines it. the intentional part is staying close to real users - having someone tell you 'that's not my problem' is worth 10x any self-study loop. the trap is more building when what you need is more friction.
That’s a strong way to put it.
I think the tricky part is that “taste” doesn’t really improve in isolation — it improves when there’s consequence.
Reps without consequence can actually reinforce bad patterns.
The fastest jumps I’ve seen usually come from:
– shipping something real
– getting clear negative signals (not vague feedback)
– then adjusting under constraint
Almost like taste isn’t built — it’s forced into shape by reality.
The hard part is most people avoid that loop because it’s uncomfortable.
Curious — have you found any specific environments or constraints that accelerate that, or is it mostly just staying close to users over time?