Thinking as talent
AI maxed every benchmark you trained for.
LeetCode, grades, the essay, the violin. A model can match them in seconds now. What still sets people apart is how they reason through a problem no one has solved. Discuria makes that visible.
This week: Why is the No. 9 bus losing riders?. No account needed to start.
Free beta · Works with the AI you already use
- LeetCode612 solved
- GPA4.0
- Violinsince age 4
- Admissions essay11 drafts
- LinkedIn3 posts a week
Even the quants
The firms that measured everything are hiring for what they couldn't.
“What remains a durable differentiator once AI has surpassed many of humanity's analytical and technical capabilities? Our answer is not what most people would expect to hear from a quantitative investor and business-school dean: creativity and emotional intelligence.”
A firm named for quantitative research now says the edge is human, and that it is hiring for it. The problem is that nothing on a resume shows it. A trace of how someone reasoned through an open problem does.
On the table now
Not just code. The questions people actually live with.
Money, health, cities, work, family, culture, ethics, technology. If it needs judgment and has no answer key, it belongs here, and so does your take on it.
- Civic & policyWhy is the No. 9 bus losing riders?
- Everyday lifeHelp a family of four cut grocery costs by a third without eating worse
- Work & careersShould a 24-year-old take the safe job or the risky one?
- CommunityA neighbourhood WhatsApp group is spreading a false crime rumour. What do you do?
- HealthYour town has one clinic and a six-week wait. Cut it in half.
- EducationA school wants to ban phones. Write the policy that actually works.
What we score
The path, not the answer.
Anyone can paste a good answer. The dead ends, the checks, the change of mind: that part is yours. The judge reads the whole trace and ties every score to a moment in it.
Example trace · Why is the No. 9 bus losing riders?
- 01
First guess: the fare is too high, so people stopped riding.
Opening hypothesis
- 02
The 2019 fare cut didn't slow the decline. It isn't price.
Humility: drops its own idea on evidence
- 03
Who left? Split riders by hour: midday is flat, the 7–9am peak fell.
Rigor: tests before building on it
- 04
The model doesn't have on-time data for this route. I'll mark that as the gap instead of guessing.
Metacognition: knows what it doesn't know
- 05
Best answer so far: reliability, not demand. This changes my mind: if on-time rates held steady.
Synthesis: a claim with a way to be wrong
How it works
01
Pick an open problem
Real questions without an answer key: everyday life, cities, money, health, work, policy, culture.
02
Think it through with your AI
Use Claude, ChatGPT or Codex through our connector. Your subscription, your keys.
03
Get your reasoning scored
Six dimensions, each tied to moments in your trace. A profile that shows how you think.
Six dimensions
A resume measures one line. People have more than one.
- Rigor
- Checks a claim against evidence before building on it.
- Creativity
- Finds an approach beyond the obvious first answer.
- Metacognition
- Notices where their own thinking is weak, and says so.
- Humility
- Lets go of a wrong idea when the evidence turns.
- Synthesis
- Joins separate pieces into one answer that holds.
- Persistence
- Stays with the hard part instead of routing around it.
Where we are
The judge is a language model scoring against a fixed rubric. It has not yet been checked against a panel of human raters; that study comes before any accuracy claim. Until then, treat a score as a beta signal, not a verdict. We'd rather tell you that than print a number we can't defend.
Our own lab has already caught a small judge giving more credit to the vocabulary of good thinking than to the thinking itself. Building a judge that can't be flattered is the work, and every score is stored with its trace so it can be checked.
Talent is evenly distributed. Opportunity isn't.