Reasoning audit — for serious writing

Check flawed reasoning before it becomes consequential.

Paste a product memo, research draft, or strategy argument. Bias Detect ranks the strongest evidence-backed reasoning risks and gives one practical next check.

New accounts start with 1,000 credits.

The bias problem

Bias hides inside confident claims.

The risky part is not the obvious mistake. It is the sentence everyone nods through because it sounds reasonable.

Claim

We should keep going because stopping feels uncomfortable.

Sunk coststopping feels uncomfortablepast effort is treated as proofWhat would we decide if this started today?
Confirmationwe should keep goingsupporting evidence is one-sidedWhat evidence would change the decision?
Availabilitythe recent win dominates the memoone vivid example becomes the base rateWhat does the broader sample say?

The instrument

Paste a serious draft.
Inspect the reasoning.

Bias Detect quotes the line, names the bias, and hands back one practical next check. Nothing leaves your machine until you run.

Specimen resultFig. 01
We should delay export controls because power users already work around it.
Bias
Normalcy Bias
Evidence
The draft treats the current workaround as a reason to defer a known user request.
Next check
Ask what would change if complaints doubled next month, then decide the launch threshold.

Full audit

The product checks the full 175-bias map, then ranks only the strongest evidence-backed matches instead of showing a generic writing score.
Full-map scan, ranked evidence

0 / 5,000 words · 0 credits

Draft saved locallyCan take a few minutes

The method

From loose claim to next move, in three readings.

  1. Paste the draft

    A roadmap, a decision memo, a research argument — anything where a quick gut check is not enough.

    Drafts stay local until you press run.

  2. Run the full scan

    We check the text against 175 canonical bias cards and rank the strongest evidence-backed fits.

    Most scans finish in under twelve seconds.

  3. Inspect the proof

    Each result returns claim, evidence, the gap, and the specific next check attached to that bias card.

    Counterevidence and missing evidence are quoted, not paraphrased.

Every dot is a candidate bias.

We score all 175 canonical cards against your draft, then surface only the strongest evidence-backed matches. The rest go quiet on purpose.

Cards scored
175
Evidence gate
met / partial / missing

Previewed bias

Overconfidence EffectNeed to act fast
  • 49Need to act fast
  • 63Not enough meaning
  • 37Too much information
  • 26What should we remember

Name the line. Quote the evidence. Give one next check.

First 1,000 credits are on the house. Drafts stay on your machine until you press run.