AI can build you a $100M startup in a weekend
What it cannot help you do is
**HELP YOU MAKE TASTEFUL CONTENT
Here is how to not make AI slop in 2026 **
Taste is the new core skill
It all began with Brockman's 5-word tweet, which went viral
- The matter of fact is that execution is commoditized, curation has become scarce.
- There is an implied trap here, the trap being that taste is some sort of a passive and static credential.
- Real taste is often just practice.
- It atrophies or it goes out of style when the person stops building reading and stops observing things
- I think that they just become their opinions, and those are very different.
The meme responses were merciless pointing to the fact
that silicon valley is full of people with strong opinions about fonts, and little else besides words and quarter-zips.
What taste is (and what kills it)
Taste is distinction under uncertainty.
It is to walk into a room of ten things and know which one is different, to know exactly why, and then articulate it in a way that changes how we see it.
Jobs had that. Rubin has that. And neither could turn it into a framework, because frameworks are for stable worlds.
Taste lives in moving ones. The moment you formalize it, you fossilize it.
Thiel lens: taste as elimination, not addition
Thiel's real insight on competition really applies here: the world often rewards the last movers, not the first. In this context:
1. The person with taste isn't always the first one to notice something is good;
1. They're often the last person to notice it so precisely and completely that nothing needs to be fundamentally added.
1. It is more about elimination of the tasteless rather than tasteless addition.
Basic Bh Taste
AI made the execution of designing, building, and shipping light speed, but the current AI models, by their definitions
give us something which I call statistical taste.
That’s my name for the default look and feel AI produces when it’s trained on massive data sets.
It is just average. It is the safest version, or the most optimal version, of good, and in 2026 safe-good is everywhere. We have seen it:
- the same purple-to-blue gradient on white-coded websites
- the same bold claim plus three bullets per CTA layout
- the polite and corporate voicing in everybody’s emails
AI did not kill taste. It made the average way more accessible.
Escaping the Mean
The real problem today isn’t AI slop anymore. The real problem is that most people will use AI and still never build taste, because they stop at the average.
- You can reach the top 75% and be “just good.”
- “Good” has become accessible to everybody.
- So every bit of delta in the next 25% matters more : the part that separates statistical taste from curated taste.
In this age, we’ll label the world in extremes:
- 75% and below -> AI slop
- Above 75% -> masterpieces
We are living in the age of extremes. This guide is about escaping the average.
Why AI Makes “Safe-Good” So Easy
AI made generic content easy because it does the three things which are required to create what I call culturally averaged aesthetics:
1. It predicts what usually works.
1. It reproduces patterns people have been rewarded for before.
1. It compresses thousands of references into one safe-good output.
That is why every first output often looks like an app store landing page, or a Medium blog post, or an average hero section on a website. It is not terrible, but it is predictable, and being predictable is the enemy of taste.
The Shift: The Average Is the New Mediocre
The real shift in 2026 is that the average is the new mediocre.
- Before AI, if something was mediocre, it was because the maker lacked the skill.
- After AI, if something is mediocre, it’s usually because the maker stopped early.
AI raised the floor, but the ceiling is higher than ever.
- Taste becomes more important because AI made the middle crowded.
- Now being above average is not enough.
Everyone can ship a 7/10.
- The only way to stand out is to become a 9/10 in judgment, a 9/10 in selection, and a 9/10 in specificity.
- And have real authentic perspective.
AI Has No Taste (That’s Why It Helps You Build Yours)
The most uncomfortable truth that people come to face is that the AI fundamentally lacks taste.
It is the greatest mirror; it will always mirror the user, because it forces a question you could avoid before:
“Why is this wrong?”
Before AI, mediocre work took time to produce. Effort always created noise that you could confuse with hard work or quality.
AI removed that noise.
Now you can generate 10 different versions of the same AI slop in 10 minutes.
The gap between what AI produces and what you actually want is where your taste lives.
Most people will always avoid that gap.
People with taste have to learn to name it.
The Core Loop: Generate, Then Destroy
This is how you use AI to build taste.
You generate, and then destroy.
- You produce 10 to 20 versions of anything.
- You don’t use it to pick the best one.
- You use it to develop a rejection vocabulary.
The muscle that you are fundamentally training is: “This fails because X.”
- X is structural, not cosmetic.
Do that enough times, and you’ve internalized a framework you couldn’t have articulated beforehand.
Find the Canon, Then Ignore the Canon
Use AI to find the canon. Then ignore the canon.
Ask AI to map the best work in any field across the last 100, 200, or 300 years.
- Read it.
- Watch it.
- Absorb it.
Use deep research not to copy it, but to understand why it worked there
Minimax-M2.5 is my personal choice :) You can use anything which works
Curiosity is the only differentiator today between average taste and curated taste.
- The more you ask these incredibly powerful models, the more you will dive deep in.
Most people are willing to be stuck at average.
Hence why they will never ask those questions to go make that curated masterpiece of software, writing, design anything.
Make With Stakes
I got inferred insight from Will Manidis's article "Against Taste"
Taste that never risks anything and stays theoretical is passive.
Publish the essay. Ship the product. Present the deck.
- The feedback loop of real output is irreplaceable.
- People respond, or don’t.
AI can accelerate everything except good old-fashioned feedback.
The lightning execution of AI has allowed us to ship and publish at light speed.
- And that has allowed us to accelerate our taste.
- The fundamental prerequisite for taste is knowing what rules you are willing to break.
And you cannot break rules which you do not know.
The Specificity Test
Every time AI generates a version, ask the question:
- Could this have been written about anything else?
If yes ; if you could swap the exact subject out and it would change absolutely nothing - it has no fundamental taste.
Taste, by its definition, is specific.
Rick Rubin’s production on *Californication * sounds like nothing else.
- Because it was made for those songs.
- At that particular moment.
- By that group of people.
By that logic, taste is irreducibly contextual.
Resistance as a Signal
When something AI produces bothers you, and you cannot understand why, don’t move on.
- Sit with that friction.
- That is the raw material of taste.
Most people will scroll past it.
But the person who stops, and interrogates, and asks themselves what is exactly wrong here is actually doing the hard work.
Over time, those cumulative moments become your aesthetic immune system:
- unexplained discomfort
- uncertainty
- lack of understanding
Taste vs “Better Than Humans” Taste
The deeper, uncomfortable truth about taste is that AI will eventually have better taste than most humans at most things.
- This is not because taste is mechanical; it isn’t.
- It’s because taste is trainable on human output.
What AI cannot replace is taste with a perspective.
Rick Rubin doesn’t just know what sounds good.
- He knows what sounds good given what he believes music is for.
- And that belief is not derivable from a corpus of data sets.
- It comes from active living.
- From choosing.
- From loss.
Taste is a product of life, and not a data set.
The answer to how you build taste using AI is to use AI to strip everything that is in taste.
- Let it produce the competent.
- The average.
- And the defensible.
And in that gap between what it produces and what you actually want is where taste lives.
The work is learning to name that gap:
- precisely
- correctly
- accurately
That’s it. That’s the whole thing for taste.
Taste Triad: Curiosity, Judgment, Standards
Your taste grows fastest when you spend most of your time sharpening your ability to see and observe.
- Curiosity (anti-inspiration): Curiosity is not liking inspiration. It is interrogating what is inspiration.
- Judgment: Judgment is the act of selection and rejection at the same time.
- Standards: Standards are the lines you won’t cross.
The truth, which hurts most people, is that if you don’t have standards, AI will supply them for you.
AI standards are culturally averaged aesthetics.
The 75–25 Rule for Building Taste Fast
A 10-Minute Daily Practice
One of the simplest daily practices that I do, for 10 minutes (if you do only one thing), is this:
1. Pick one output (a paragraph, a landing page, or a tweet) and generate 10 versions with AI.
1. Write one critique line for each version using the phrase “fails because…”
1. Rewrite one version using one constraint (for example: no adjectives, no gradients, one idea, depending on the medium).
1. Publish, because that loop will build faster than more inspiration.
The Final Shift
Before AI, what took you 10 hours could be mediocre work that got passed off as quality because of the sheer hard work it required.
With the advent of AI, the process has fundamentally become easier.
Now no longer can hard process be used as an excuse for mediocre quality and taste matters more than ever.
“Do not let hard process excuse bad results.” -Sam Altman
