Research programme
Human Foresight Lab
Could you have seen it coming?
People love guessing what happens next. Who did it. Whether it works. Whether it lasts.
Prediction platforms record the guess. We record the reasoning behind it.
One engine, many domains. The same question everywhere: what do you think will happen, and why?
Working prototype
Answer, explain, then see what you missed.
Episode 1 of 6 · 47 minutes watched
Who committed the murder?
0 words · at least four
Held back until you commit
The outcome stays hidden. So does everyone else's answer.
00:09:41
John's watch is eleven minutes fast in the hallway shot.
00:21:03
Alice's coat is soaked, though it stopped raining at six.
00:38:17
The study window is locked from the inside.
00:44:52
A second set of keys is missing from the board.
Every clue is timestamped. What you use is recorded.
Prediction isn't the moat. Reasoning is.
Most systems store a yes or a no. That tells you who was lucky.
We store the sentence underneath it — the evidence used, the evidence ignored, the confidence attached.
Over years that becomes a record of how people actually think, and where thinking reliably fails.
Outcome
Right or wrong. The least interesting column.
Reasoning quality
Did the stated reason actually support the answer?
Evidence used
Which available clues appear in the sentence.
Evidence ignored
What was on screen, or on the page, and went unused.
Calibration
Does 80% confidence mean right about 80% of the time?
Earliness
How little information you needed to get there.
One engine, many domains
The modes
Mystery Mode
Drama & crime fiction
Pause at the end of an episode. Name the culprit, then justify it in one sentence.
Episode 1 of a six-part series: Alice, John, or someone not yet on screen?
Measures · Clues used, red herrings avoided, how early you were right.
Poirot Mode
Timestamped clue tracking
Every clue in the episode carries a timestamp. You mark the moment you suspected.
"I think it's the secretary" — logged at 00:14:22, before the second body.
Measures · Which clues you saw, which three you walked past, who else spotted them.
History Mode
Before you read what happened
Take the decision as it was faced, without hindsight, then compare with historians.
Would Caesar cross the Rubicon? Would Napoleon march on Russia?
Measures · Your reasoning against the historiography, not just the outcome.
Science Mode
Long-horizon technology
Predict a technology's trajectory, write down why, and get called back years later.
Will atmospheric water harvesting become mainstream by 2035?
Measures · Calibration over a decade — including the forecasts you'd rather forget.
Museum Mode
Standing in front of the object
Guess how the work was received in its own time, then discover what critics wrote.
Would critics of the day have called this Turner revolutionary?
Measures · How well you model a period you didn't live in.
Field Mode
Inside our own prototypes
The same engine runs behind Quinn Labs products, recording why, not just what.
One Fleet: why did the driver override the advice? One Waste: was this item really recycled?
Measures · Where the model was wrong and where the human was wrong.
Not a betting product
No stakes, no odds, no money. Nothing here pays out.
The reward is finding out whether your intuition holds up, and reading how someone else reached a better answer from the same evidence.