When AI is wrong, will you notice?

Undumb is a 10-minute daily workout to practice independent judgement and exercise the parts of your brain that need it most.

see the exercises
Focus
Reason
Recall
9:41

Wednesday

Good morning
Today’s mind workout 10 min

Today’s focus

Deep reading + evidence

Focus

74

Reason

82

Recall

68

Read closely. Challenge the narrative

Domain: real-world decisions Skill: Argument evaluation
Undumb reasoning set 1 of 5

An online marketplace piloted a ranking model that promoted unfamiliar sellers. Buyers exposed to the model purchased from a wider range of sellers and returned more often 30 days later. The company concluded that seller diversity increased retention.

However, the pilot was offered only to buyers who had previously browsed at least five seller pages. The treatment also shortened the delivery estimates shown on product cards.

Which additional finding would most seriously weaken the company’s interpretation?

Strong reasoning.

That result removes both plausible confounders and eliminates the observed lift, directly weakening the claim that seller diversity caused retention.

Games and tests to reclaim your brain.

Find the next move when the rhythm changes.

The arrow and dot count follow separate patterns. Choose the next tile.

Two patterns, one answer. Pick a tile.

Compress without flattening the truth.

Source draft

Although the beta performed better on engagement, we should probably avoid launching it globally right now because the experiment was short, the sample excluded new users, and support contacts increased.

≤ 45 wordsLead with the decisionName the tradeoff
0 / 45 words
One strong answer

Hold the global launch. The beta raised engagement, but the short test excluded new users and increased support contacts. Extend the experiment to a representative cohort before deciding whether the engagement gain justifies the operational cost.

Measure progress on the cognitive functions that matter to you.

Daily practice adapts to your goals.

Example report

Your cognitive profile

Last 12 weeks
3tracked areas

Deep focus 74

Reasoning 82

Recall 68

Session consistencySteady practice
8 sessions
90756045
MayJunJul
This month’s insight

You reason well with short passages. Accuracy drops when context exceeds 900 words.

Next week adds two longer reading sessions while keeping the subject matter familiar.

Next checkup August 12

18 minutes · 5 skill areas

When AI is wrong, will you notice?

What are you looking to improve?

Choose all that apply. We'll tailor your daily workout around it.

Almost there.

Add your email to lock in early access.

Thanks for helping shape Undumb.

Your spot is saved. We’ll be in touch as Undumb takes shape.