the deepest space is within

a lap timer
for your mind.

Baseline your brain in under ten minutes, then watch how it moves — across sleep, medication, caffeine, seasons, years. You, compared to you. No population averages, no leaderboard.

cognogram.org look in, not up 0 light-years away
the idea
Cognition, recorded.

A -gram is a written trace. A cardiogram of the heart. An encephalogram of the brain. Cognogram is one of the mind at work.

Eight short, well-worn tests take a reading of how your brain is doing today — across processing speed, sustained attention, impulse control, working memory and executive flexibility — and turn it into a number you can watch over time. The only benchmark that matters is your own last lap.

One baseline. Then every lap after it.

what we measure

Five windows onto the doing brain.

Eight tests, arranged into the five domains they measure. Not a theory of the mind — just an honest map of what these particular tasks can see.

Processing speed
How fast the eyes-to-thumb wiring fires, and how quickly you scan and match under the clock.
Reaction Time · Symbol Sprint
Sustained attention
Staying on task through something dull — and how steady you are minute to minute.
Go / No-Go
Impulse control
Your brakes. Whether you can withhold the response you already started making.
Go / No-Go
Working memory
Holding things in mind, then working with them rather than just holding on.
Digit Vault
Executive flexibility
Switching rules without losing your place, and filtering out what doesn’t matter.
Color Clash · Connect · Arrows · Switchboard
why it's different
Built to answer "how am I changing?"
01

You vs. you

Most tests rank you against a population. Cognogram compares you to your own baseline — statistically powerful, and honest about what it can claim.

02

Short enough to repeat

The full battery is under ten minutes; a single check-in is sixty seconds. Run it rested before a med change, or groggy on four hours of sleep — it picks up the difference either way.

03

On the record

Every lap is saved and dated. Watch a trajectory build — then hand your clinician a clean report, not a hunch.

04

Shows its own error bars

Where a result is uncertain, it says so and prints the range. Where two distributions overlap, it draws the overlap. Every method and limitation is published in full.

play it, don't study it
Three games. Each one is a real cognitive test underneath.

Most brain-training apps wrap an invented task in a game. We did the reverse — took paradigms that already exist in the published literature and changed only the scoring. No account needed to play.

Steady

simple reaction time

Not how fast you are — how consistent. The target band tightens every time you land inside it.

measures → reaction-time variability, the most replicated cognitive marker in ADHD

Dead Reckoning

time reproduction

A light stays on, then goes out. Hold for exactly as long as it was lit. No counting.

measures → your internal clock — and it's the one game a faster device can't win

Hold Fire

go / no-go

Hit every green. Never hit a red. Red shows up a quarter of the time — which is what makes it hard.

measures → impulse control, separated from inattention
Why the scoring looks odd. A gaming monitor and mouse beat an older phone by roughly 40 ms on any reaction test — more than a sleepless night costs you. So a leaderboard on raw speed ranks equipment. We score spread, differences, and seconds-scale tasks instead: the measures your hardware can't inflate.
open the arcade  →
beyond the baseline
Tests measure you. Experiments ask questions.

Once there is a baseline to compare against, the same eight tests can be pointed at specific, falsifiable questions. Each of these states plainly what it can and cannot conclude.

⚡ Ignition

Can a browser test catch a drug?

Run a vigilance task repeatedly across one session while a model tracks how much caffeine is actually in your bloodstream, minute by minute.

why caffeine
It blocks adenosine — the same system vigilance performance runs on. Well-characterised kinetics, legal, cheap, already in the house.
why within-session
Sampling repeatedly against a curve is worth roughly 3–5× comparing two separate sessions.
the honest limit
Falsifiable by design: if the model says peak and nothing moves, the instrument failed its own test.
◎ Stimulant Response

Is my medication working?

On-medication runs against your own clean off-medication runs, matched for time of day. An effect size instead of an impression.

why variability
It is the test index most reliably sensitive to stimulant versus placebo — more than speed, more than errors.
why it's needed
Stimulants reliably make people feel sharper whether or not performance moved. Prefrontal signalling follows an inverted U: more is not better.
the honest limit
Cannot select, titrate or monitor a medication. Treatment often improves without normalising, so a still-elevated score is not a failure.
⚖ Convergence

Should I get evaluated for ADHD?

A validated screener alongside attention measured at two timescales and under two levels of arousal — because how your attention moves says more than where it sits.

why converge
A computerised attention test alone tops out near AUC 0.78. Self-report and performance correlate only weakly, so those two blocks carry genuinely different information.
what it returns
A probability updated from a stated base rate, with correlated channels down-weighted and an honest uncertainty band.
the honest limit
At a ~4.5% base rate, a full positive still lands near 45–50% — “worth an evaluation”, not an answer. Structurally better at lowering the estimate than raising it.
⚘ The Long Night

Does creatine protect a tired brain?

Two matched short nights — one without creatine, one with. The comparison is the change from your own evening baseline on each.

why this one
In rested adults creatine does approximately nothing to cognition. Under sleep loss, a single dose improved reaction-time dispersion by about 9% — which is precisely what we measure.
why matched nights
A rested baseline cannot test creatine. Only a sleep-deprived night without it can.
the honest limit
Both published acute trials come from one laboratory, and neither enrolled people already taking it daily.
Every paradigm, citation and limitation is written up in full — including the pharmacology of caffeine and ADHD medication. Read the method →
the open roadmap
What this can pin down — and what it can't.

Most self-tracking tools ship a measurement, imply it means something, and never say what it could not have detected. Here is the arithmetic in the other direction. Every question is a ratio: the size of an effect over the size of your own noise. The numerator gets all the attention; the denominator decides whether any of it is knowable.

what you're trying to detecteffectsessions
A sleepless night+32%~5
One bad night (~5 h)+16%~12
Stimulant, clear responder−22%~13
Caffeine, true washout−22%~13
Caffeine, habitual morning cup−12%~37
Creatine, acute, sleep-deprived+9.2%~61
Creatine, chronic, rested~1%>500
Matched sessions needed at conventional confidence, on reaction-time variability. Green is comfortably in reach; red is a study nobody finishes alone.

The split that matters

Those counts are for one person. Sixty paired sessions is an impossible ask alone and an entirely ordinary study across thirty people doing two nights each.

The individual question is hard. The population question is easy. Conflating the two is the standard error in this whole category — consumer tools claim population findings from an individual's data, and researchers dismiss self-tracking because the individual case is weak. Both are looking at the wrong axis.

Sleep loss is the clear winner here, and it isn't close: two to four times larger than anything else on the list. If Cognogram is good at one thing, it is telling you what a bad night costs you.

questions we could actually answer
Not aspirations. Specific, falsifiable, and currently open.
~30 people
Does creatine do anything for a sleep-deprived brain outside one lab?The entire acute literature comes from a single group. Independent replication is the cheapest valuable thing on this list.
~400 people
Does an acute dose still work if you already take it daily?Every published trial enrolled non-users. Millions take it daily and take extra after a bad night, on a theory with no data behind it in either direction.
~1,000 people
Does day-to-day variability predict ADHD better than a single session?A clinic sees one day. Whether between-day inconsistency adds anything on top is unknown — and belongs to whoever collects the data first.
~10,000 people
Who responds, and who doesn't?Sex, diet, age and metabolism all appear to moderate these effects. Moderators need roughly ten times more people than main effects, which is why almost none are established.
And where more people stops helping. Somewhere around ten thousand, statistical power stops being the constraint and bias takes over — self-selection, unverified self-report, unmatched conditions. Past that, more people do not buy more truth. Which is why the roadmap front-loads reliability and reference-standard work rather than recruitment: the ceiling here is measurement quality, not sample size.
what would actually help

The bottleneck is sessions, not ideas.

Every question above is limited by one thing: how many people run a consistent protocol and keep running it. A single session contributes almost nothing. Ten matched sessions from one person contributes a usable noise estimate — the denominator this entire page is currently forced to assume.

If you take creatine, drink coffee, take a stimulant, or simply sleep badly and want to know what it costs you: the most useful thing you can do is establish a baseline and be boring about it. Same time of day, same device, repeatedly. That is unglamorous, and it is the entire input this needs.

Research preview. Not a medical device, not a diagnostic instrument. Data stays on your device unless you choose otherwise; contribution to any aggregate is opt-in, off by default, and no collection has begun — that requires ethical review and a published governance policy first.

Map your mind.
Look in, not up.

start your first lap  →
read the science →