Skip to main content

intuition is taste

· 4 min read
Anurag Atulya

— Anurag Atulya

We're entering an era where AI agents write code proposals, pitch decks, articles, architecture docs — all of it. Fast. Plausible. Confident.

And that's exactly why taste is what will matter — in engineering and in business, for founders and for teams, technical or not.

Not taste as in "I prefer minimal UI." Taste as in: this is good, this is bad, this will survive, this will rot — before you can fully explain why.

That feeling has a name people use interchangeably: intuition. I think intuition is taste. Same thing. Gut feel earned the hard way.

you don't get it from a prompt​

Taste doesn't download. It doesn't arrive on day one of using an agent.

It comes from doing one thing — or many related things — for a very long time. So long that it's repetitive. Boring, even. You ship the same class of problem again and again until you know it in and out.

Code: you've seen the handler that should've been a pure function a hundred times. You feel it in the first diff.

Business: you've pitched, lost, won, and watched which stories land before the spreadsheet proves it.

Writing: you know which paragraph is throat-clearing and which sentence actually moves someone.

That's not magic. It's pattern recognition compressed into a reflex. Intuition = taste.

The agent can generate ten options. Taste is knowing which nine to kill without a meeting.

good taste and bad taste​

Founders need both — and I mean that literally.

You develop taste by doing things so many times that you build intuition for what's good and what's bad:

  • Good taste: the abstraction that still makes sense when the team doubles
  • Bad taste: the shortcut you recognize because you were the person who shipped it at 2am and paid for it for two years

You need sensitivity to bad outcomes, not just good ones. Senior judgment is often "I've seen this movie." That's taste stored as scar tissue.

If you only ever accept agent output because it compiles or sounds smart, you never train that muscle. You outsource the repetition — and the intuition never forms.

what matters when generation is free​

When agents write code proposals, pitch articles, system designs, hiring emails — the bottleneck stops being can we produce this?

It becomes:

  • Is the logic sound, or just fluent?
  • Is this code we'd own, or debt wearing a confident comment?
  • Is this design right for our scale, or a blog post architecture?
  • Is this pitch true to what we actually build?

Technical and non-technical work converge on the same question: does this person — or this org — have taste?

Volume won't save you. More tokens isn't more judgment. Scraping Slack into a fine-tune doesn't encode how your best people think. It encodes Slack.

why we built Taste Studio​

I built Atulya around memory — retain, recall, reflect — because agents need somewhere to learn from experience.

Data Forge exports what memory can prove: citations, timelines, grounded rows.

Taste Studio is the other lane: what you choose to teach. The examples you'd defend. The review comment you'd want every junior to internalize. The pitch paragraph that sounds like you, not like everyone else's LLM.

Same idea as intuition in a person:

  • Seed — the original judgment (immutable)
  • Working copy — iterate without losing what you started from
  • Diff — see what changed and whether you still trust it
  • Variants — expand from a golden example without diluting it
  • Retain — put curated taste back into memory so the agent lives it, not just trains on it once

Mechanics: Taste Studio guide. This post is the why.

a note for founders​

My advice, blunt:

Do the thing. Code, sales, product, hiring, writing — pick the loops that matter for your company and run them until you're bored of them. That's when taste starts.

Use agents to accelerate output, not to skip the repetitions that build intuition.

When something feels off and you can't articulate why yet — pay attention. That's taste trying to speak. Write it down. Turn it into a seed example. That's how personal judgment becomes organizational judgment.

The next era won't reward who generates the most. It'll reward who selects the best — who curates taste and compounds it.

That's the moat. Everything else is commodity.

further reading​

— Anurag Atulya