A daily practice in discernment

Anyone can get an answer now. Learn to tell if it's a good one.

Honecraft is a two-minute daily practice that trains judgment, the thing a chatbot can imitate the sound of but doesn't actually have, one real object or real decision at a time.

Get the app drop See how it works

Visual & spatial · today's practice

Johannes Vermeer, View of Houses in Delft, known as The Little Street

What is this doing to your eye before you've named a single brushstroke?

Guessing

Certain

Johannes Vermeer, The Little Street, c. 1658. Rijksmuseum, public domain. You commit before you find out why it works.

A glimpse of the archive

Vincent van Gogh, Farm in Provence — Van Gogh
Farm in Provence

Payag, Shah Jahan on Horseback, folio from the Shah Jahan Album — Payag
Shah Jahan on Horseback

Utagawa Hiroshige, Six Jewel Rivers from Various Provinces — Hiroshige
Six Jewel Rivers

Kitagawa Utamaro, Mother and her Children in Front of a Freestanding Screen — Utamaro
Mother and Children

Alphonse Mucha, portrait of Leslie Carter — Mucha
Leslie Carter

Asante-Akan artist, Lion Ornament — Asante-Akan
Lion Ornament

Henri de Toulouse-Lautrec, The Sofa — Toulouse-Lautrec
The Sofa

Utagawa Kuniyoshi, Sumo Wrestlers — Kuniyoshi
Sumo Wrestlers

“

This tells me something true about how I think that I could not have gotten by asking an AI chatbot.

The one sentence every part of Honecraft is built to earn

How it works

One practice, five minutes, five moves.

See something real

A painting, a passage, a business decision, a claim. Sometimes a convincing fake sits right next to it.

Commit, out loud

Choose, and say how sure you are, before you get one more fact. No hedging, no do-over.

Learn what you missed

The reveal names the principle you probably didn't have a word for, and argues the other side of it too.

Take the view you skipped

You're handed the angle you underweighted and asked, seriously, whether it changes your answer.

Watch the pattern, not a score

Over weeks, a private Mirror: what you notice, what you actually value, how well your confidence matches reality.

The Mirror

Evidence about how you think, not a score.

What you notice first

Illustrative. Yours builds from your own responses.

Composition

People

Systems

Evidence

RestrainedExpressive

Moderate confidence, based on 41 responses

Nothing shows up until there's enough evidence to be honest about it. No percentile rank, no personality type, no “top 3%.” A confidence band on every claim, because a number without its uncertainty is a guess dressed up as a fact.

Every line in the Mirror traces back to the specific responses that produced it. You can always ask why it thinks what it thinks.

Once a month it goes further: which of your first answers changed after the reveal, and what that says about where your judgment is actually moving, not just where it sits today.

Evidence trace, response #14: you rated Botticelli's The Annunciation high on Composition, low on Systems. That's the kind of line every claim in your Mirror can point back to.

Inside the app

Four places. Nothing else.

Today

The day's practice. Nothing else on the screen.

Explore

Paths, focused labs, the full stimulus archive, things you've saved.

Mirror

What you see, value, choose, and how well-calibrated you are.

You

Challenges sent and answered, history, account.

What you'll practice on

Six domains, rotating.

Ground truth

Detect the AI

A photo, a paragraph, a logo, a bar of music. Real, or generated. Refreshed as the models get better.

Taste, no right answer

Visual & spatial

Painting, poster, typography, product form, textile. Pulled from public collections, not stock libraries.

Taste, no right answer

Built environment

Buildings, streets, civic design. The hardest category to source honestly, and worth it.

Partial ground truth

Decisions & trade-offs

Real business, civic, and policy calls, with the actual outcome where one exists.

Ground truth

Evidence & claims

Statistics, source quality, base rates. The muscle that keeps the rest of the practice honest.

Taste, no right answer

Ethics & stakeholders

Genuine dilemmas with no clean answer, where the reveal is the whole point.

Why this works

Leonardo da Vinci, A Bear Walking, red chalk sketch

Leonardo da Vinci, A Bear Walking, c. 1480

Experts and novices aren't looking at different things. They're sorting the same thing differently.

Show a novice a physics problem and they'll group it with other problems that look alike, the ones with an inclined plane in the picture. Show an expert the same set and they'll group it by the principle underneath, this one's really about conservation of energy, next to a problem that looks nothing like it on the surface.

That shift, from what something looks like to what it actually is, is discernment. It breaks into three moves: notice the structure that's actually there, work out which of it matters and to whom, then commit to a call under real uncertainty and stay willing to revise it. Honecraft trains those three, in that order, on purpose, every day.

Who it's for

People who already lean on AI daily, and are a little uneasy about what it's doing to their own thinking.

Designers, strategists, writers, product people, anyone whose job is judgment calls and not just output. And the friend they send a hard question to on a random Tuesday, who opens a text, answers in ninety seconds with no app to install, and gets curious enough to come back.

V1 pricing

Free, permanently. Paid, if you want more of it.

Free

One complete practice a day, forever, full quality, full reveal
Unlimited challenges sent and answered
Your Mirror, as the evidence supports it
30-day practice history
A restrained monthly Mirror card

Studio

$11.99 / mo

Unlimited practice, any day
The complete Mirror, every spectrum, full evidence trail
Longitudinal view of how your judgment is changing
Bring Your Own Decision: run a real decision of yours through the same practice. It never tells you the answer, it gives you a record of your own reasoning.
Unlimited Detect-the-AI, refreshed weekly

$79/year for the first 1,000 people, locked for life. No countdown timers, no leaderboard, no “top 3%.” A product about judgment that monetized badly would be lying about what it's for.