# People want to get better: a philosophy, and a route

> An essay by Matt Edmund (Warsaw, October 2026). What he believes about AI, judgment and getting better; the vision (digital twin coaches, the job market, Tacit Mind for everyone); and the route: Tacit Mind as the foundation, Drop The Hands as step one, coaching integration as step two, coaches beyond the platform as step three. The interactive version: https://33prime.io/vision.html

Most of them don't know how, and they're embarrassed to ask. This is what I believe about that, and the route we're taking to do something about it.

## Part one: the philosophy

### 01. Most people are just trying to be a better version of themselves.

I came up through telecoms and recruitment, spent twenty years in Europe, and co-founded a software company that went from four people to about two hundred engineers. If all of that taught me one thing about people, it's this. Almost nobody wakes up wanting to be replaced. They want to be better at the thing they care about: a better golfer, a better engineer, a better parent.

What stops them usually isn't talent. It's the cost of asking. Asking a person costs money, it costs their time, and it costs a little pride. So people don't ask. They hit the same slice for three years. They don't ask the senior engineer the thing they think they should already know. They search at eleven at night and get a hundred confident answers that don't fit them.

That's what AI genuinely changes. Not the person. The cost of asking. If you can ask something that knows how you think, at any hour, without anyone watching, you'll ask more. And people who ask more get better faster. The person stays the lead. The AI just makes the asking easy.

I said the line above in April, in one of 337 conversations I kept while building a twin of myself. It's the one I keep coming back to.

### 02. Answers are free now. Judgment isn't.

For most of history, answers were scarce. You went to school to collect them and paid experts to hand them over. Most of how we learn, hire and promote is still built around that scarcity.

That world is gone. Answers are instant and free, and plenty of them are wrong in ways that sound right. I call it the Walmart of information: aisles of things that look like they work, and you freeze in front of them. People aren't short of answers. They're drowning in them. The future skill is the questions, not the answers.

So learning shifts, from doing to grading. You do something once, properly, so you understand how it fits together. After that you mostly judge the work instead of doing it. That isn't a lower bar. You have to know a thing well to catch a mistake in it. Learn enough to tell good from bad.

### 03. The best judgment in the world is stuck in someone's calendar.

If answers are cheap, judgment is what's scarce. It's what a great coach has that a textbook doesn't: knowing which of your twenty problems to fix first, what to say, and what to leave alone today because you've had enough.

Right now that judgment is rationed by the hour. A good coach might teach thirty hours a week. Everything between lessons, which is most of a student's life, happens without him. His method doesn't travel. It waits for the next booking.

I think that's the most important thing AI can fix, and the most important rule while fixing it is this: the person keeps the human part. Fear, motivation, and the moment someone's about to quit. Nobody's job disappears. The best part of it travels further.

### 04. It's the in-between that matters, the why.

Agents are the easy part. Every company will have them, and they'll be cheap and interchangeable. What makes an AI yours isn't the model. It's your reasons: why you decide the way you do, what you care about, what you'd never do.

So the layers settle like this. Humans manage human things: relationships, trust, the room. Your twin manages the agents: which one, what counts as done, what you'd have said. The agents do the tasks.

And one rule keeps it safe. Intent stays home. Your twin goes out and reads every catalog. Nobody comes in and reads your twin. Your reasons, your health, your family and your money stay at home. What leaves is a question. What comes back gets judged.

## What we're for

Help people become better versions of themselves, by putting the best human judgment within reach of anyone who wants to get better, with AI that answers to the person, not the other way round.

## Part two: the vision

### Three futures, one idea

If people want to get better, and judgment is the scarce thing, then the job is to let the best human judgment travel while the person stays the lead and private things stay private. That one idea plays out in three places.

Each is a different market. They share one contract, and one foundation underneath.

### Digital twin coaches

Every coach, in every sport and then beyond sport, has a twin that carries their method to every student between sessions. Students find the coach whose method fits how they learn, through whatever AI they already use. The coach spends their time on the moments only a person can handle.

- Coach profiles and plans, owned by the coach
- Twins measured against the coach's own calls
- Discovery by fit, not by advertising

### The job market

AI-native recruitment. Candidates and companies both have twins. They meet in a room between them, compare what each has actually done and needs, and only the real fits reach the people. No cold messages, no six-second CV scans, and roles you'd never have searched for.

- Receipts from real work, not self-written profiles
- Twin-to-twin matching before any human talks
- Neither side's private intent leaves home

### People: Tacit Mind

Everyone gets the thing I'm building for myself first: a twin that knows your reasons, manages your agents the way you would, and keeps your intent at home. You stay on the human side: the relationships, the trust, the room. It handles the rest, and earns more of it over time.

- One memory behind every AI you use
- Autonomy earned one kind of decision at a time
- Proven first on me: a real interview on 30 January

## Part three: how we get there

### Built from the ground up

Underneath everything is Tacit Mind, the foundation: one memory behind every AI, the reasons behind every decision, private rooms that stay private, and intent that stays home. Every product we build stands on it.

On top of it we build in steps, starting somewhere small, real and checkable. The first step happens to be golf. Not because the vision is about golf, but because golf is the most honest place we know to prove the contract before it goes anywhere else.

### Step one, on Tacit Mind (live today): Drop The Hands

The first thing we've built on the foundation is a golf app, and that's on purpose. Golf has the one thing most AI products don't: a ground truth. A model can say anything. A golf ball can't. The shot went where the caddie said it would, or it didn't, and you find out in seconds.

So in Drop The Hands, the maths decides and the model only explains. Thousands of simulated shots pick the aim and the club; the language model turns that into a sentence, and a guard checks it didn't make anything up. The engine gives the same answer to the bit on a phone and on a server. Golf Course Atlas turns real courses into geometry the engine can trust.

It also does the first thing the philosophy asks for: golfers get better without having to ask anyone. Every round adds to what it knows about how they play.

How:

- A deterministic engine in Rust, the same on the phone and the server, checked against a golden set of shots.

- Monte Carlo strokes-gained for aim; a rule-based policy for club choice.

- Course geometry from Golf Course Atlas, with a validator that rejects anything physically impossible.

- A contradiction guard between the numbers and the words.

For the vision: Proof that an AI can carry judgment without making things up, plus real players and real coaches to build the next layer with.

### Step two (building now): coaching integration, the coach inside the caddie

The caddie is good. A great coach is better. So step two brings the coach in. In coach mode, the caddie becomes the coach's twin: his method, his cues, his drills and the order he fixes things in, learned from his own lessons and his own verdicts on his students' swings.

We don't onboard a coach with a form. The twin reads what he's already produced, then reads it back to him: here's what I think you believe about the swing. He says yes, no, or "true, but not for the twin". Every answer makes it more his.

- **11:04 PM.** A student pulls it left again. Too embarrassed to text her coach this late, she asks his twin. It answers in his method: quiet hands this month, try his towel drill before Saturday.

- **7:30 AM.** One line waits for the coach, not a transcript. Her misses moved left over three rounds. Your call.

- **7:32 AM.** He changes her focus. Then he spends his morning on the two students who were about to quit.

The rule that matters most: the twin never overrides the coach. If it notices something, it tells him privately, and he decides. That isn't a line in a prompt a clever question can talk around. It's built into the system. And it works the other way too: players find the coach whose method fits how they learn.

How:

- A coach profile: his philosophy, cues, vocabulary, the faults he fixes first, his drills. Owned by him, versioned.

- A plan per student that only the coach can write: the current focus, the swing thought, the drills.

- Learning from his verdicts. Before he labels a swing, the twin predicts his call. Its hit rate is the measure of how much it sounds like him; each miss becomes a rule he approves.

- Earned autonomy. Drill suggestions run on their own once it agrees with him often enough. Swing changes always go to him.

For the vision: The first digital twin coaches: a person's judgment reaching every student between lessons, with the person still the lead.

### Step three (next): coaches beyond our walls

A coach's method shouldn't live inside one app, including ours. Step three takes the coach's twin outside Drop The Hands, on Tacit Mind, and lets it go wherever his students already are. In Claude, in ChatGPT, on his own website, inside another app, through one door.

His method travels with him, not with us. He isn't locked into our platform, and that's deliberate. If he leaves, his twin leaves with him. His private notes stay private: his twin goes out, and nobody comes in.

How:

- One memory, one door. The twin lives in the coach's own Tacit Mind and connects to any AI through a standard connector, so Claude or ChatGPT can ask it directly.

- A room per audience. Students, other coaches and the public each see only the layer meant for them. Private notes never leave.

- Portable by design. The coach owns the record, the method and the history. We host it; he keeps it.

For the vision: The twin belongs to the person, not the platform. That's what lets the same foundation serve coaches, the job market and everyone else.

## The principles we hold ourselves to

1. **The person stays the lead.** AI extends people. Autonomy is earned one kind of decision at a time, never assumed.
2. **The engine decides, the model explains.** Anything that has to be right is computed and checked. The language model explains it; it doesn't get to make it up.
3. **Receipts over claims.** What you've done, with the reasons, beats what anyone says about themselves. Including us.
4. **Intent stays home.** Your twin goes out; nobody comes in. Private things stay private, and there's nothing to sell.
5. **Prove it on ourselves first.** My own twin sits a real interview against me on 30 January. If it can't pass as me, it isn't ready for anyone else.

## Tell me where I'm wrong

Some of this will turn out to be wrong. That's fine, and it's why I wrote it down. The parts you push back on are the parts I care most about.

Matt, Warsaw, October 2026 · matt@33prime.io