In sixth grade (mid-1980’s), I won an award at school for my project about the state of Earth’s drinking water. What I had researched back then (with as much “research” as a sixth grader could do prior to the internet, which basically amounted to World Book Encyclopedia and a few newspaper microfiche) led to the inevitable conclusion, reinforced by numerous news articles, that humanity would be out of drinking water by the year 2000. It was a young age to acknowledge the sobering conclusion that I had a front seat to witnessing the final decade of humanity’s existence in the universe.
This led to some amount of existential agita, but perhaps not as much as you’d imagine. I was, after all, only in sixth grade. The fact that several kids in my class could afford Omega Supreme was mind-blowing enough to occupy the majority of my mental cycles.
Remember when, two years ago, journalists couldn’t stop sounding the death knell of software development, declaring “prompt engineering” as the durable profession of the future? This seemed an inevitable conclusion of simple extrapolation: engineers are typing more prompts → getting prompts right is hard → the future will be inherited by prompt experts.
Who even talks about “prompt engineering” now? In fact, who even admits we were wrong about that seemingly bulletproof Nostradamus extrapolation?
We were right that prompts would play an increasingly large, if not primary or exclusive, role in software development. What we completely ignored — which, in hindsight, is glaringly obvious — is that AI would get smart enough to obviate deft prompting as a prerequisite to producing useful code. We saw a direction vector, ignored everything else, and drew a line to infinity.
There’s nothing more natural than linearly extrapolating emergent trends to the nearest conclusion they collide with. It’s much harder, even for me, a sixth-grade project prize winner, to acknowledge we’re all still drinking potable water here in 2026. In fact, if I’m honest, I’ve never even bothered to follow up on what happened. I just assume some analog to the Green Revolution happened in hydrology.
The latest such future-of-software prognosticating is all about “taste.” Now that we’ve suddenly agreed (without any acknowledgement or fanfare at all) that “prompt engineering” is an outmoded skill, we’ve latched on to the next frontier of difficulty, which is deciding what exactly to build.
I think this is misguided for two reasons.
It’s not that the word “taste” has no meaning. It’s more that its meaning may be as useful (or not) as “AGI.”
Have you noticed that talk about whether AGI will happen (or even has happened) has nearly disappeared, much in the same vein as there being little discussion about the fact we’ve blown clear past the Turing test?
The reason for this is definitional.
For some people, “AGI” referred to the ability of AI to displace human intellectual labor. We’re clearly at a point now where AI is able to substitute for some traditional human labors (e.g. the majority of a paralegal’s job), so referring to this question of labor substitution as “AGI” is no longer useful.
For others, “AGI” was always a moniker that meant “whatever ineffable magic essence that’s uniquely attributable to humans.” From this perspective, “AGI” has always, by explicit construction, been a moving goalpost.
What — your computer can beat chess? I bet it can’t beat Go, because only humans can solve that incomputable problem with their indomitable spirits.
Wait — you’ve now solved Go? You’ll never be able to create art.
Your art and music are now beloved by millions of Americans? It’s derivative and will never be truly creative.
“Taste” as a term has this exact same problem. For some number of people, “taste” merely means whatever the current AI isn’t able to do for you. Sure, it can jam out code, but does it know the best UI? What, it can now create good UI? Well, it certainly doesn’t know what apps people truly desire. “Taste” becomes shorthand for “what action to take next.” It’s essentially synonymous with “what prompt to issue.” As long as AI isn’t initiating 100% of all problem solving itself, these people will call that phenomenon a lack of “taste.”
As with “AGI,” the other half of people using the word “taste” really secretly mean, probably without even being conscious of it, “an ineffable quality unique to human beings which computers couldn’t possibly ever have.” When these people say that the frontier of human-piloted AI is all about “taste,” what they actually mean is, “Don’t worry, your judgment will still be valuable, because after all, you’re human, and that means you have abilities which no amount of silicon could ever possibly have, because, you know, because!” These are the same people who become confused and distressed when confronted with the fact you can already buy rack-mountable compute powered by human neurons.
Let’s suppose I haven’t convinced you that “taste” has become as useless a moniker as “AGI.” Let’s instead suppose what you mean by taste is what Steve Jobs dismissively referred to when naming the biggest thing he felt Bill Gates lacked.
This sense of the word is what I hear most from PMs who loudly declare the future is All About Taste™. You can sense the relish with which they anticipate the rise of a key faculty essential to their profession, the one axis that traditionally differentiated the strongest PMs from engineers who “merely” code. Deciding what to do is, after all, far more important than deciding how to do it.
I grant this is completely true. If you have the taste of Steve Jobs, I readily acknowledge you will inherit the Earth. The rise of AI can only serve to make you even more powerful and far more valuable than you’ve ever been, as each barrier between idea and delivery gets ever shortened and eliminated.
But here’s the problem: every black-turtleneck-wearing PM I’ve ever worked with thinks they’re Steve Jobs. And, to a T, none of them are.
Not because they aren’t skilled. In fact, I’ve worked alongside many talented PMs in my career. Instead, what’s happened is that the term “Steve Jobs” has undergone the same conflation with “unparalleled runaway commercial success” that “taste” is about to (if it hasn’t already).
If you’re not using the word “taste” to mean whatever quintessence humans have which machines never will, you’re likely using the word as shorthand for “whatever will make this product loved by the masses.” And from that perspective, you probably don’t have it. At least not better than AI.
The question is not whether there exist a handful of humans who have exceptional taste. The question is whether the average person, or even the far-above-average person, does. You see, the hard part about taste is nearly everyone thinks they have it. But quite demonstrably, empirically, nearly nobody does.
Did the creators of Beanie Babies have taste? Tickle Me Elmo? What’s Jony Ive doing these days? What about all the people who read Creativity, Inc or any other book about Pixar? Are you now seeing a boatload of movies undergirded by outstanding story arcs?
I am not saying that nobody has it. What I’m instead saying are two things:
We should be careful lest the word “taste” come to embody an ambiguous amalgam of “preferences initiated by a human” and “product decisions correlated with commercial success.” The word will then have lost its utility.
Even if we were to be very careful to use the word “taste” meaningfully, odds are incredibly good you don’t have it. Just imagine the average PM you’ve worked with in your career and what they thought of their own sense of taste. You are, more likely than not, exactly that. At least I am.
We are all the Dunning Kruger of creative papillae.
The next time someone tells you that the future of software engineering is all about “taste,” protect yourself from the false sense of security which the term apparently conveys to many people.
They either mean, “Don’t worry, do the thing that computers can’t yet do,” which feels tautological, or they mean, “Emit prompts that build products universally loved,” which is as useful as counseling startup founders with the profound business insight that they should “build something people value that sells for a profit.”
On one hand, exactly correct. On the other, impracticable.
On our way out the door to go ice skating, my well-meaning dad once gave this parting advice to my brother and me: “Don’t fall.”
I now tell you this: “Have taste.”