Taste Is Necessary, But Not Sufficient

tl;dr: to be able to successfully use taste to select ideas, you need good ideas. A few people can generate new good ideas. Its not clear that AI can.

In Which Pete Creates New Things

I once worked with a CTO/co-founder called Pete. Most of the time Pete did good CTO things: he built hard parts of the system, helped more junior engineers and influenced company strategy. But now and again Pete would go away for a few days, and come back with a prototype of something completely new. The rest of us had been diligently doing the right things: collecting customer feedback, prioritizing feature lists, making tradeoffs. But Pete would come back with a thing that no customer had asked for, none of us had considered, and nonetheless was just obviously right.

"I think Henry Ford once said, "If I'd asked customers what they wanted, they would have told me, 'A faster horse!'" People don't know what they want until you show it to them”   — Steve Jobs

A tech company needs process, feedback, structure, careful market consideration and all the rest. But it also needs moments of brilliance, moments when something completely new enters the frame, shakes things up, creates new possibilities.

Which brings us to creativity, and the loud conversation about taste.

Where Do New Ideas Come From?

“Imagination is more important than knowledge. Knowledge is limited. Imagination encircles the world.”  — Albert Einstein

“Creativity is just connecting things” — Steve Jobs

A good general model of creativity goes like this: first, a generative phase of research, brainstorming and ideation during which, somehow, the spark occurs. And then an editing or synthesis phase of adding, subtracting, refining and finally finishing. 

Taste is all about the second phase. Taste selects, refines and edits. The first phase, the sudden unexpected arrival of something new and brilliant, is much more mysterious.

I’ve been in countless strategy and brainstorming meetings with very senior, highly intelligent people where the whiteboard ends up covered in ideas, all of them smart and sensible, all of them selected with taste, but none of them really alive, none of them right. And I’ve known a handful of people who, every once in a while, would show up with a prototype, or a sketch, or a sentence and it was clear, immediately, that this was the basis for the next few years of work. These were the first phase people, the generators.

Steve Jobs was a master of the second phase, a synthesizer and editor. He had the ability to notice the sparks that were emerging, and connect them to the right people. And then give the people the resources and narrative drive the build the thing that the sparks were suggesting. (I’m enjoying these posts from Stan Ng about the creation of the iPod).

Steve Jobs had (brilliant) taste.

Taste vs Brilliance

“The universe buries strange jewels deep within us all, and then stands back to see if we can find them.” ― Elizabeth Gilbert, Big Magic: Creative Living Beyond Fear

"The most beautiful thing we can experience is the mysterious. It is the source of all true art and science." — Albert Einstein

(I’m aware that Albert Einstein and Elizabth Gilbert make an odd juxtaposition :-)

What’s missing, I think, in the current conversation is that taste is necessary but not sufficient. To build a great company you need moments of chaotic, creative brilliance. Without brilliance, taste is simply doing the next thing until the next disruptive company or technology destroys you.

You need moments where somebody (and often it's one person) has an extraordinary idea, a spark. Sometimes these are small, but powerful: reimagining an interface layout, reworking a positioning. Sometimes that are prototypes that appear from nowhere and define the next few years of your company's growth.

These were sudden, new possibilities. Taste dosnt generate those possibilities. Human brilliance does.

An early model of creativity published in 1926 has four phases, which looks helpful until you notice that phase three is “Illumination”: “the appearance of the ‘happy idea’…”. In other words, you prepare yourself, put yourself in a state in which you can be inspired, and then, inspiration just kind of happens.

What this misses is out is that the model originated from Hermann von Helmholtz, an absurdly talented and creative nineteenth century physician and physicist. He had taste, certainly. But he was also clearly one of the minority - the brilliance generators. The creativity model devolves to: put Hermann in a room and wait.

Taste recognizes the ideas, nurtures them, gives them room to grow, but doesn’t generate them.

Rick Rubin’s The Creative Act describes the work of taste specifically:

“The editor’s role is to gather and sift. Amplifying what’s vital and whittling away the excess. Culling the work down to the best version of itself… Editing is a demonstration of taste.. Our taste is revealed in how our work is curated”

(Rick Rubin, like Steve Jobs, is a synthesizer, an editor. Like Steve Jobs, he has excellent taste).

Can AI Replace Human Brilliance?

In an agentic org we employ humans, with taste, to manage agents. Which begs the question: where does the unexpected brilliance come from? What are the limits of AI here?

I don’t know. And neither do you :-)

The skeptics would say that LLMs are fundamentally backward looking and are incapable of genuine novelty. Boosters would say that examples like the recent breakthroughs in mathematics and the famous Move 37 by which Alpha Go beat a human Go grandmaster show that they can do something awfully close. Skeptics would push back by saying both examples just use a lot of clever brute force. Boosters would say there are reasons to believe they are doing something like conceptual thinking (see also Anthropic's recent paper).

We don’t know. We’ll be finding out.

Right now, evidence suggests that asking AI to “be creative” produces a very large amount of homogenous, tasteful gloop.

In a recent experiment, LLMs helped authors write short stories by producing ideas. Individually the authors got more creative — they had quick access to a pile of novel (to them) themes. But collectively the stories began to resemble each other.

Which aligns with the recent posts about design slop: you vibe code an app. It looks great. But it also looks like every other vibe-coded app. Your taste was … fine. But you were choosing from a large range of generic options.

Rick Rubin warns us:

"Rules direct us to average behaviors. If we’re aiming to create works that are exceptional, most rules don’t apply. Average is nothing to aspire to”

What To Do

1. Add “brilliance” to your vocabulary in addition to “taste”. Without brilliance taste becomes, over time, traffic management, a series of endless course corrections to a path resembling all other paths.
2. In addition to the managers of agents, you need people in the org who have this trait. The ability to look at a problem or a road map or a design or a system structure and see it differently, go away for a while (it can be five minutes) and come back with something new.
3. You need to recognize, protect and nurture these people. A lot of the time they may not incredibly productive, may not fit into the new world of super-managers. Keep them close anyway. Naming is good. Call them something special.
4. When the genuinely new idea shows up, it will probably look wrong, or strange, or unacceptable, particularly in a world where agents are busy shovelling up hundreds of carefully modulated, perfectly reasonable, color inside the lines ideas. Google famously missed that they had invented the single most disruptive technology of the past fifty years. Xerox famously created the Macintosh and MS Windows (and Ethernet etc etc), but didnt realize it. You may need some courage to grab hold of the new thing and nurture it until its strong enough to hold its own.

We’re all making up the new agentic organization as we go along. Tech companies are high-stakes creative acts. Its important that we recognize that the essence of creativity, if not its execution, remains stubbornly human, at least for now.

Interested in coaching as we move through this amazing period? Hit me up!

A Few Links (Mostly About Creativity)

Olivia Moore at A16Z discovers that her autonomous X-posting AI can create all day, but none of it is interesting.

Simon Sinovsky on Harvard banning personal computers back in the day. Because: new technology! Mental corrosion!

Saanya Ojha on a Brown University professor discovering that take-home exams no longer work. Because: new technology! Mental corrosion!  (If universities can't teach taste, who's going to corral the bots?)

Can AI Have Taste? A replit vibe-con wrestles with the issue. Kind of. A fireside chat was titled:“Everyone’s a creative. Now what?” As I write above - not true - some have taste, a small few have brilliance.

Highly creative people have different brains.

Favoring the 40 year old founder. Because they have experience, and therefore, yes, taste. Its a theory, and I'm here for it.

Narrative Watch

In an age of narrative whiplash, it’s interesting to watch the tides whip in and out.

Agents are the new consumers. A16Z now funds startups that make stuff agents, not humans, want. (See also the Salesforce “headless” news a few weeks ago). 

Models are just commodities. This has been coming for a little while. Model routers anybody?

How I Use AI In These Posts

I use AI for research. I don’t use generated text or editing in these posts.