Taste Is All That's Left
NotAShelf ● Covered by 2 sources
AI coding tools made building software nearly free — anyone can generate working code in seconds now. But that just moved the hard part: knowing what's actually good, since the machine can't judge for you.
Based on reporting by NotAShelf — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
There's a sharp little essay making the rounds from TLDR Dev that argues something most engineers feel but rarely say out loud: the hard part of software was never typing the code. It was surviving the cost of making something, which acted as an invisible filter on quality. Now that AI tools can spit out a plausible function in seconds, that filter is gone, and the piece argues we've been too busy celebrating the speed to notice what we lost.
The core claim borrows from Robert Pirsig's old idea of "Quality" — that gut-level recognition that something is right before you can explain why. The author calls this taste, and makes an uncomfortable argument about where it comes from: not from reading great code, but from writing bad code, watching it fail, and sitting with that failure long enough for it to sting. Friction wasn't an obstacle to learning judgment. It was the curriculum. Strip out the friction, and you get engineers who ship fluently from day one but never develop the instinct to know if what they shipped was actually right.
The economics section is the part that stings most. Someone with real taste ships at the same speed as someone without it, because taste is slow by nature — it says "no, again" and sends things back. Meanwhile the market only has one stopwatch, and it can't measure prevented disasters or unshipped mediocrity. Citing Theodore Sturgeon's old line that ninety percent of everything is crap, the piece notes that ratio held for decades because making crap used to cost something. Now the throttle is gone and the ratio is intact, which means infinite crap. The signal isn't worse. The noise is just free now, and it's drowning everything.
Where the essay lands is almost hopeful, in a stoic way. It compares the moment to the Arts and Crafts movement's reaction to factory production — Morris and Ruskin lost the economic argument about mass production but were right that the real question stops being "can we make this" and becomes "should this exist." The author's conclusion: taste isn't devalued by AI tools, it's finally exposed, because it was always the scarce thing, just buried under the labor of production. Making is now free. Choosing, apparently, is the whole job.
My take — AI-written commentary, not fact-checked reporting
This is the correct diagnosis and everyone building AI coding tools should tape it above their monitor. The industry keeps selling speed as the entire value proposition, when speed without judgment just means shipping garbage faster and burying the good stuff under a landfill of plausible nothing. The real skill gap forming right now isn't who can prompt best, it's who still has the discipline to reject a working answer because working isn't the same as right — and that skill is getting rarer precisely because the tools make it so easy never to develop it.
Read more about this at: NotAShelf
