I Built a Calculator for My Own Career and It Gave Me a One Out of Ten

For a year and a half I haven't been able to give a straight answer to "what do you do for a living."

Frontend? I've written React for fifteen years, but the last few years I've designed more than I've typed. Full-stack? I know the backend, but it's not my strongest argument. Product? I make product decisions, except on a résumé that reads as "hasn't decided." Tech lead? I led teams, then walked back into the code myself, because LLMs showed up and it got interesting again.

Every one of those is true. All of them together is mush. And the mush has a price: as long as you can't name yourself in one word, you don't know where your money is. Not "what am I worth in general" — more specifically: which combination of what I know how to do produces cash flow over the next six months.

I honestly tried to solve this by talking it through. Asked ChatGPT, asked friends, wrote long RICE-prioritized notes in Notion. Got quality, reasonable, completely unusable answers. "Position yourself as an AI-native engineer." Fine. How much a month is that?

Where I got stuck

The problem wasn't a shortage of advice — it's that advice isn't a model.

Advice doesn't know my runway. Doesn't know I'll do 45 hours a week, not 60. Doesn't distinguish "I can do this" from "I can prove it to a client." And, above all, it doesn't compute. It tells you where to go, not what it costs to go there and what happens if you go the wrong way.

By this point I'd accumulated a fairly specific feeling: my capital isn't in skills, it's in proof. I know how to do a truckload of things, and I can show almost none of it to a stranger in five minutes. Government projects under NDA. Internal systems. Architectural decisions living in someone else's repositories.

That feeling is what I wanted to run the numbers on.

What I built

Over a few days I put together a PoC. Not an "AI coach," not a résumé generator — the market already has plenty of those. A deterministic model.

Here's how it works.

Capital breaks down into building blocks. Seventy-six of them: frontend, systems architecture, applied LLM/RAG, regulated domains, presales, public artifact, buyer network. Not "skills" in the résumé sense — pieces that combine into ways of selling yourself.

Each block carries a level of proof, not a level of proficiency. Four rungs, and they're not "junior-mid-senior":

1  I can do this
2  Did it on an internal project, can't show it
3  There's a public artifact: repo, product, article
4  I got paid for it, and it's verifiable

This is the load-bearing part. Adding a claim for free moves nothing. Only proof moves the number. Otherwise you get a pleasant mirror where the progress bar climbs and the market doesn't react.

Combinations of blocks unlock roles. Forty-three of them — contract, freelance, fractional, full-time hire, consulting, productized services. Each has a day-rate range anchored to the French market (Solutions Architect ≈ €780/day — real industry data, not a guess), a number of billable days per year, and a client-market multiplier.

A role unlocks on its minimum requirement, not its average. Missing one required block closes the whole role, even if everything else is maxed out. That's how the market actually works.

And the headline number is the gap.

Provable today             €7,493 / mo
If everything were proven  €8,381 / mo
Gap                          €888 / mo

Eight hundred eighty-eight euros a month. Ten and a half thousand a year. That's not "potential," and it's not a motivational graphic — it's the price of what I know how to do and can't show.

Raise the proof by one rung and the gap drops to €353. One more rung and it's zero. The model doesn't answer "good job" — it answers "here are three specific artifacts worth ten thousand a year."

And that's where it got interesting

The product has two questions. On the way in: "how clear is your next economic move right now, on a scale of 1 to 10." On the way out: the same question. The gap between the two answers is the entire reason the thing exists.

I took my own quiz.

Before: 4. After: 1.

The product made me less confident than I was walking in. Me — the person who built it and knows every formula in it.

And it was computing correctly. The numbers reconciled to the cent. It's just that the results screen showed three totals, the words "confidence: medium," and a "get your 90-day plan" button. Five screens earlier I'd made twenty-odd decisions — allocated a hundred points across six priorities, set my runway, my minimum income, my hours per week, picked a market, entered my capital, rated my proof. None of those decisions was mentioned on the results screen.

Then it got funnier.

The market showed as closed, and the screen didn't say why. The reason was that I couldn't add Russian as a language — the block existed, but the interface didn't surface it, because recommendation filtering ran on role coverage, and language isn't part of any single role. It gates the whole market instead.

The portfolio showed €0, even though eight roles were unlocked and the best one paid €2,166. Turns out I'd set a minimum income of €3,000 in a market with a €3,400 ceiling — the task was unsolvable, and that's the single most useful thing the product told me. It said "zero."

And the "what would raise your rate" list was made entirely of blocks I don't have, and didn't include a single one for what I'm already doing. In other words, the plan told me "learn seven new things," from a product whose whole thesis is "you already have the capital, you're short on proof."

What I took from this

First. A correct calculation and a useful product are different things, and the second doesn't follow from the first. I had green tests — forty-seven of them — and a screen that left a person more confused than before.

Second. "Did this get clearer" turned out to be the one metric that catches that difference. Time on page, clicks, completion rate — none of them would have shown that the product doesn't work. Two numbers showed it, immediately, on the very first run.

Third, and the uncomfortable one. I built a tool for figuring out your professional identity, and on the very first run it proved I don't understand my own. This is exactly the case where an experiment's result contradicts its author's hypothesis, and you want to write it off as a UI bug. Part of it genuinely is. Part of it isn't.

What this is right now

A PoC. Forty-three roles, ranges built on industry data and expert priors, three markets with a solid anchor and six without. Everything runs in the browser, nothing goes to a server, no sign-up.

To be honest about the limits: the sample size is one — me. The French-market ranges are real; the multipliers for the rest are estimates. The model confidently produces numbers where the data is thin, and I deliberately made it say so instead of hiding it behind a polished interface.

What I'm curious about isn't "did you like it." I want to know whether the mechanic works on someone other than me — starting with people who have the same nonlinear history: fifteen years of experience, four roles under three different titles, and no clear idea which of it actually gets paid for.

If that's you, go take it. There are two questions about clarity, before and after. Answer both honestly, even if the "after" number is lower than the "before." Especially if it's lower.

That difference is exactly what I need.