# Who to learn from

> How to pick the few people worth learning from when anyone can sound like an expert, using one test (have they shipped something people use?) and two tricks for finding more.

Module: Working with AI · Level: beginner · 6 min · https://school.agents.org.in/foundations/working/who-to-learn-from

Your feed has eleven threads this week on the only design prompt you'll ever need. They're all confident and all well written, and a couple of them disagree with each other. You have one free evening. Which one deserves it?

## What you read, you become

The things you take in shape how you think, slowly and without asking permission. Read a hundred posts that say every screen needs a gradient and one day you'll reach for a gradient without deciding to.

That's always been true. What's new is that a model can now make anyone sound like an expert. Fluent writing used to be a rough sign that someone had spent years on a subject. Today it's free. So choosing whose ideas you let in has become a skill of its own.

## The test: proof of work

There's one question that cuts through almost all of it: have they shipped something people use or admire? A product with real users, a library other projects depend on, a typeface millions of people read every day. Call it proof of work (Something a person has shipped that others use or admire: a product, a library, a typeface. It's the best reason to trust their advice, far better than followers), something a person has made that others rely on. It's the best reason there is to trust their advice.

If you want to learn type design, read the person who drew a typeface you see on every phone, not the most-shared thread about font pairing. If you want to learn interface motion, it helps to know that Emil Kowalski's advice on it comes from building Sonner and Vaul, a toast library and a drawer that a great many apps run on.

There's a second, weaker test for when you can't see the work: do they work somewhere that's famously hard to get into? Getting there took proof most people never see. If someone passes neither test, they may still be fun to read. **It's probably just good posting.**

*[Interactive demo: Three invented people. Walk each one through the two questions.]*

## Followers measure posting

Followers, likes and viral threads measure one thing well: how good someone is at posting. That's a real skill, and some people have it alongside a lot of shipped work. But building and posting are different skills, and it's easy to mistake one for the other when the posting is very good.

*(A handy check: could you name one thing this person made that you'd miss if it disappeared?)*

## The invisible makers

Many of the best builders never post at all. They're busy. The people who designed your phone's settings screens or your favourite app's onboarding may never write a word about it.

You can still learn from them, because their work is public even when they aren't. Use what they made closely, as if you'd been asked to explain it. The next time you open your phone's settings, try naming three decisions someone made: why the search sits where it does, why that toggle doesn't ask for confirmation, why the list is grouped the way it is. Putting a decision into words is how you take it in.

> **Ask your agent to read a page with you.** Point your agent at a screen you admire and ask it to list the decisions it can see, with a reason for each. Then argue with its list. You'll notice more than either of you would alone.

## Find more, keep few

Once you've found one person who passes the test, look at who they follow, quote and admire. Good people tend to point at other good people, so this narrows the search fast.

Then keep the list short, five to ten names. They don't all have to be designers. A sushi chef who has spent thirty years on one craft has a lot to teach about restraint. Finally, let your agent help you take their ideas in, always with the sources attached so you can check what it tells you.

**Find people with proof of work (paste into your agent):**

```text
I want to learn interface animation from people with proof of work.
List eight people who have shipped something widely used in that area (a library, a product, a well-known site), with the thing they shipped and a link to their own writing.
Mark any claim you couldn't verify. No one whose main output is threads.
```

## What to remember

- Before you take someone's advice, ask what they've shipped that people use. Followers measure posting, not building.
- Learn from people who never post by using their work closely and naming the decisions in it.
- Find more good sources through who your best one follows, and keep the list to five or ten.

**Check yourself:** Which is the strongest reason to take someone's advice on building date pickers?

- [ ] They post about dates every day
- [x] They maintain a date picker thousands of projects depend on
- [ ] A model summarised their thread
- [ ] They have the most followers in the topic

*Why:* A date picker thousands of projects depend on is proof of work: strangers trust it with their product. Posting often, a summary or a big following says nothing about whether they can build one.

Go deeper: [Turn what you know into a skill](/foundations/working/knowledge-into-skills) · [Taste is trained](/learn/philosophy/taste-is-trained)

## In the skill

- https://school.agents.org.in/skill/meta/reading-others-interfaces.md
- https://school.agents.org.in/skill/philosophy/taste-is-trained.md