Black Gates
A rhythm dodge, and why it is shaped this way
Predictability as Scaffolding
The beat is anticipable rather than reactive. The Daily is a single deterministic seeded board shared by every player that day, so its arcs recur, become memorisable, and reactive dodging gives way to rehearsed sequence timed against a known pulse. Progress indexes mastery, not variance. Nothing in the field is noise.
Agency Without Stakes
The gap-set poses a genuine decision, which sector and how soon, while stripping out the cost of resolving it wrongly. Absent an enemy, failure is not defeat by anything that intended it; you lose to gravity, an impersonal condition. Current learned-helplessness theory inverts the original: passivity is the unlearned default under sustained aversive exposure, and the variable actually acquired is the detection of control. Failure that is always legible supplies that signal at no cost, repeatedly.
What You See Is What Kills You
Drawn position and judged position are identical, because nothing pulses spatially and the two cannot drift. Rings never twin, since two walls arriving on one beat would generate an invisible second gap-set. The beat clock halts on tab-blur, otherwise every ring in flight resolves simultaneously on return. Arrival timing is validated against a simulated player carrying explicit decode-and-react cost rather than estimated. Every death is attributable, and you always know why.
Equal Under Gravity
Nothing carries between runs. No progression, no accrued difficulty, nothing owed. We all start equal and return to the same resting state, which is to say there is no challenge-load. Allostatic-load theory locates the cost of stress not in the response but in its failure to terminate and reset, so a structure that resets completely each cycle is the inverse shape.
An Exit Rather Than a Hook
Level 40 is a win with an ending, which most arcade games lack entirely, because failure is what sells replays. One deterministic board per day, unrerollable. No streaks, no login rewards, no notifications, no account, and therefore no data. It is built to absorb for roughly five minutes and then release. ADHD makes cessation costlier than initiation, so the game supplies the stopping-cue rather than delegating it to depleted executive function.
Modes as a Regulation Menu
Six, and not difficulty tiers. Classic wants clean anticipation. Ascension inverts gravity so rings rise and the whole field defamiliarises. Charted climbs a ladder of polygons. Zendless discards scoring entirely, and death carries no consequence, because you return as the same ball and begin again. The operative question is which demand your head can hold today, not how much punishment it will absorb.
The Floor
One sector closes every eight levels, floored at three open, so half the ring remains passable at any depth. It cannot degenerate into a needle. That is a hard constraint, not a difficulty curve.
On Timing
The researchers who built the rhythm game used throughout the clinical literature published, separately, that commercial music games cannot train rhythmic skill because their input windows are too coarse, measuring roughly 100 ms in Guitar Hero. This runs a fixed 8.33 ms tick, twelve times tighter, tick-clocked rather than wall-clocked so behaviour is identical at 60 Hz and 240 Hz, with every input logged at tick resolution so that any run re-simulates exactly.
What It Is Not
The one game to reach a regulatory clearance for ADHD moved a surrogate attention score; its symptom endpoints did not separate from sham. A review of 26 reviews of digital therapeutics found under a third report adverse effects at all, and game addiction sits among those they name. That is the harm this is built against.
It has never been tested. It treats nothing and manages nothing. The mechanisms above are real research; the artefact is a bet on them, made by someone who wanted it to exist.
Maier & Seligman, Psychological Review 123(4), 2016 · McEwen, NEJM 338(3), 1998 · Kollins et al., Lancet Digital Health 2(4), 2020 · Gabarron et al., BMC Psychiatry 25(1), 2025 · Bégel et al., Front Hum Neurosci 11, 2017 · Wibble & Pansell, Exp Brain Res 241(7), 2023 · Williams et al., Ear & Hearing 42(5), 2021
Six Invariants
And the reason for each
Draw equals collision
The ring you are looking at is the ring that judges you. Nothing pulses spatially, only in brightness, so drawn and judged position cannot diverge.
Rings never twin
Two walls arriving on one beat would produce a second gap-set the player cannot perceive, so that state is unreachable by construction rather than filtered at runtime.
The clock halts on blur
Hidden tab, paused beat. Otherwise every ring in flight resolves the instant you return, which kills you for closing a laptop.
Arrival stays inside human latency
Gate timing is verified against a simulated player carrying explicit decode-and-react cost, making the tuning sim-backed rather than estimated.
The field never closes past three sectors
One closes every eight levels, floored at three open, so half the ring stays passable indefinitely. The ramp was deliberately slowed fourfold after an earlier tuning collapsed the field into an unreadable annulus by level three.
8.33 ms fixed tick, tick-clocked
Frame-rate independent and deterministic, with every input logged at tick resolution so any run re-simulates exactly. The team behind the rhythm game used in the clinical trials published that off-the-shelf music games cannot train synchronisation because their windows are too coarse, roughly 100 ms in Guitar Hero. This is twelve times tighter than the bar they said those games fail.
Why
All six exist for one reason: every death must be attributable. Legible failure is the whole design, because current learned-helplessness theory holds passivity to be the default under sustained pressure and control-detection to be the acquired variable.
The rest follows. Nothing carries between runs, so everyone starts equal and returns to the same resting state. Level 40 is a win with an ending. One deterministic board per day, unrerollable. No account, no streaks, no notifications, and therefore no data. Zendless discards scoring, and death carries no consequence, because you return as the same ball and begin again.
It has never been tested and it treats nothing.
Maier & Seligman, Psychological Review 123(4), 2016 · McEwen, NEJM 338(3), 1998 · Kollins et al., Lancet Digital Health 2(4), 2020 · Gabarron et al., BMC Psychiatry 25(1), 2025 · Bégel et al., Front Hum Neurosci 11, 2017 · Wibble & Pansell, Exp Brain Res 241(7), 2023 · Williams et al., Ear & Hearing 42(5), 2021
What Is Not In It
The design, by subtraction
No enemy
Rings close because rings close. You lose to gravity, an impersonal condition, not to something that intended it.
No arbitrary death
Rings never twin. The clock halts on tab-blur. The ring you can see is the ring that judges you, so you always know why.
No progression
Nothing carries between runs. We all start equal and return to the same resting state, which is to say there is no challenge-load.
No needle
The field never closes past three open sectors, so half the ring stays passable at any depth.
No hook
Level 40 is a win with an ending. One board per day, unrerollable. No streaks, no notifications, no account, and therefore no data.
No stakes in Zendless
Scoring is discarded entirely and death carries no consequence, because you return as the same ball and begin again.
No claim
It has never been tested and it treats nothing.
What is in it
A pulse you can anticipate, arcs that become memorisable, and about five minutes.
Maier & Seligman, Psychological Review 123(4), 2016 · McEwen, NEJM 338(3), 1998 · Kollins et al., Lancet Digital Health 2(4), 2020 · Gabarron et al., BMC Psychiatry 25(1), 2025 · Bégel et al., Front Hum Neurosci 11, 2017 · Wibble & Pansell, Exp Brain Res 241(7), 2023 · Williams et al., Ear & Hearing 42(5), 2021
Pick What Your Head Wants Today
Six modes, not six difficulties
A rhythm dodge in the browser. Rings close on the beat, six gaps, they do not line up. Roll into an open one before the ring lands.
Something you can anticipate → Classic.
The pulse is a cadence rather than a surprise, and the arcs recur until you know them.
The same problem everyone else is having → Daily Descent.
One deterministic seeded board, identical for every player that day. Mastery rather than variance.
It to stop making sense for a minute → Ascension.
Gravity inverts, rings rise from the hole, the field defamiliarises without becoming crueller.
Structure to climb → Charted.
A ladder of polygons from ring to twelve-sided, across forty levels.
None of it to matter → Zendless.
No scoring. Death carries no consequence, because you return as the same ball and begin again.
Whichever you pick
These are not difficulty tiers but distinct demands, and on some days exactly one of them is the one you can hold.
Nothing carries between runs, so you start equal and finish at the same resting state. Nothing arbitrary takes you, and you always know why. Level 40 is a win with an ending. No account, no streaks, no notifications. It is built to be put down after roughly five minutes, because ADHD makes cessation costlier than initiation and the game should be the thing supplying that cue.
It has never been tested and it treats nothing.
Maier & Seligman, Psychological Review 123(4), 2016 · McEwen, NEJM 338(3), 1998 · Kollins et al., Lancet Digital Health 2(4), 2020 · Gabarron et al., BMC Psychiatry 25(1), 2025 · Bégel et al., Front Hum Neurosci 11, 2017 · Wibble & Pansell, Exp Brain Res 241(7), 2023 · Williams et al., Ear & Hearing 42(5), 2021
In Short
Nine words and what follows each
Gravity
No enemy. Nothing in it wants anything from you.
Legible
Rings never twin, the clock halts on tab-blur, and the ring you can see is the ring that judges you.
Equal
Nothing carries between runs, so it is the same start and the same resting state every cycle. No challenge-load.
Floored
Never closes past three open sectors, so it cannot degenerate into a needle.
Learnable
The arcs recur and become memorisable, rehearsed sequence rather than pure reaction.
Finite
Level 40 ends. One board per day. No streaks, no account, no data.
Weightless
Zendless discards scoring and death carries no consequence. Same ball, begin again.
Exact
An 8.33 ms tick, tick-clocked, every input logged. Twelve times tighter than the window the researchers said commercial music games fail at.
Honest
It has never been tested and it treats nothing.
Maier & Seligman, Psychological Review 123(4), 2016 · McEwen, NEJM 338(3), 1998 · Kollins et al., Lancet Digital Health 2(4), 2020 · Gabarron et al., BMC Psychiatry 25(1), 2025 · Bégel et al., Front Hum Neurosci 11, 2017 · Wibble & Pansell, Exp Brain Res 241(7), 2023 · Williams et al., Ear & Hearing 42(5), 2021
For Clinicians and Researchers
Design rationale, evidence tiers, and explicit non-claims
This is a design-rationale document. No efficacy claim is made anywhere in it, and the artefact has never been evaluated. What follows states which mechanisms the design is anchored to, at what evidential tier, and where the anchoring stops.
Controllability, not helplessness
The design leans on the 2016 reformulation rather than the 1967 original. Passivity under prolonged aversive exposure is the unlearned default, mediated by dorsal-raphe serotonergic activity; what is learned is the detection of behavioural control, gated ventromedially, which then inhibits that default. The design correlate is a set of hard invariants guaranteeing that every failure is attributable, so the controllability signal is delivered continuously rather than intermittently. Tier: strong for the mechanism. No inference is drawn from it about symptoms.
Allostatic load: mechanism, not index
The invocation is deliberately narrow. The framework's taxonomy separates failure to habituate to a repeated stressor from prolonged response through delayed shut-down, and the design maps onto both by discarding all inter-run state, so no challenge-load accrues across cycles. The measurement literature is a different matter: there is no consensus operationalisation, fidelity to the original stress construct is frequently absent, and a 2023 individual-participant meta-analysis across 67,126 people exists precisely because standardisation does not. The mechanism is invoked; the index is not, and no allostatic-load score is claimed, measured or implied.
Sensorimotor synchronisation: near transfer only
The core demand is a motor response phase-locked to a perceived beat, and because the arcs are deterministic and recur, execution shifts from paced synchronisation toward memory-paced continuation. The trainability evidence is real and narrow. In randomised trials with matched active-control games:
composite beat-tracking p=.015, η²=0.02 · paced tapping to music p=.013, η²=0.014 · paced tapping to metronome p=.67 (null) · rhythm perception p=.16 (null) · n=26, 13 per arm
The gain appeared on the measure most proximal to the trained task and vanished on the two most distal. All four trials in this literature originate from one laboratory whose senior author declares a commercial interest in rhythm-based intervention, and no independent replication exists. Tier: preliminary. Note also that this is a spatial dodge rather than a tapping paradigm, so transfer from this input topology to the measured one is an inference, not a finding.
The far-transfer boundary, stated rather than skirted
Trained tasks improve; generalisation beyond them is weak or absent. Working-memory training yields g≈0.80 on the trained task against g≈0.05 for nonverbal ability. Balance training yields SMD 0.79 on trained tasks against −0.07 to 0.18 on untrained balance tasks, so specificity holds even within a motor domain. Accordingly no claim is made regarding executive function, symptomatology, social cognition, proprioceptive acuity or daily functioning.
Negative findings acknowledged up front
The largest trial of improvisational music therapy in autism, 364 children across nine countries with blinded assessment, found no effect on symptom severity, ADOS social-affect mean difference 0.06, 95% CI −0.70 to 0.81. Meta-analysis of music training finds cognitive benefit collapsing to approximately zero once design quality is controlled. In the ADHD digital-therapeutics literature, blinded outcomes reduce ADHD-total effects to SMD ≈0.12, and the residual inattention effect doubles when measured in the setting where training occurred, which is the signature of setting-specific practice rather than treatment.
Accommodation rationale
Per-channel audio trims rather than a master mute, because decreased sound tolerance decomposes into at least hyperacusis, misophonia and phonophobia, with meta-analytic current hyperacusis prevalence in autism at 41.4% (95% CrI 37.2–45.8) and inter-individual variance in which sounds are aversive large enough that a fixed setting cannot serve the population. Tri-state motion control with the OS preference as default rather than as a ceiling, because visual-motion hypersensitivity is documented in vestibular disorder, concussion and migraine, symptom expression fluctuates day to day, and the pause menu is reachable mid-run. Autism and ADHD appear in that literature at single-study level only and the systematic review declines to draw conclusions, so they are not claimed. Orthostatic-intolerance populations are not represented in it at all, so no visual-motion claim is made for them either.
Photosensitivity, measured
The failure flash was assessed rather than assumed: a 2.52% relative-luminance change against the 10% general-flash threshold, red ratio 0.718 against the 0.800 saturated-red bar, single decaying event rather than a periodic strobe, and suppressed entirely under reduced-motion. The rotating radial pattern remains the open question, since spatial frequency is a pattern criterion that a motion setting does not address, and its measured contrast falls well below the relevant luminance limb.
Non-claims, exhaustively
Not evaluated. Not an intervention. Not a therapeutic, adjunctive or diagnostic instrument. No claim of symptom change, functional improvement, transfer, regulation, or benefit of any kind in any population. The design is anchored to mechanism literature; that anchoring is not evidence of effect, and the distinction is the entire point of this panel.
If something here is wrong, or overstated, the feedback form is anonymous and I would rather know.
Maier & Seligman, Psychological Review 123(4), 2016 · McEwen, NEJM 338(3), 1998 · McEwen, Ann NY Acad Sci 840, 1998 · Guidi et al., Psychother Psychosom 90(1), 2021 · Johnson et al., Soc Sci Med 192, 2017 · McCrory et al., Psychoneuroendocrinology 153, 2023 · Jamey et al., Ann NY Acad Sci 1558(1), 2026 · Bégel et al., Front Hum Neurosci 11, 2017 · Melby-Lervåg, Redick & Hulme, Perspect Psychol Sci 11(4), 2016 · Kümmel et al., Sports Med 46(9), 2016 · Bieleninik et al., JAMA 318(6), 2017 · Sala & Gobet, Mem Cognit 48(8), 2020 · Cortese et al., JAACAP 54(3), 2015 · Westwood et al., Mol Psychiatry 28(4), 2023 · Kollins et al., Lancet Digital Health 2(4), 2020 · Gabarron et al., BMC Psychiatry 25(1), 2025 · Williams, Suzman & Woynaroski, Ear & Hearing 42(5), 2021 · Wibble & Pansell, Exp Brain Res 241(7), 2023 · Fisher et al., Epilepsia 67(2), 2026