Home / Mario

Super Mario Bros. · solo from the screen

One Cadence brain beat all of World 1 in a single 5-life gauntlet

Brain W1b ("the balanced finisher", trained overnight 10/31) cleared 1-1, 1-2, 1-3 and 1-4 (Bowser) back to back with the same weights, losing three lives on 1-3, then ran out on 2-1, a level it never trained on. On 50 held-out starts per level it clears 1-1 7, 1-2 8, 1-3 9 and 1-4 27 times: every level, none consistently yet. Other brains still do better on single levels (1-1 50/50 with W1_F2ck110, 1-4 50/50). Every button in every clip is the brain's.

Latest (11/16): Fabius (recall memory plus prediction slots) does not beat Cato2, the same brain without them: 920.8 against 958.0 mean x at chunks 81 to 100, a gap of -37 inside the 65 between Cato2's own two seeds. With two seeds per brain they can't be told apart; if anything the point estimate leans toward Cato2. Details. Scipio (11/14 to 11/15), a slow memory aimed at the stalls where Mario holds one button against a wall, did not change the stalls (the read). A likely reason, found after the read: a tie rule in the teaching labelled most of the 4-2 start stalls, and the 2-2 stall we checked, as correct. 4-4 is a different case. The next brain, Marcellus, started 11/16 to test this (details). In the public gauntlet two Scipio brains cleared 1-1 at chunk 50; none has passed 1-2. The goal is World 8, the final world: never coached, learned by own play only, cleared with at most 1 death in total. The goal.

W1b's gauntlet, with its brain. The same nine attempts, re-recorded with the brain's real activity next to the game: what it sees, its body sense, its neurons and the button it presses. All nine replays matched the gauntlet log exactly (end x, decisions, outcome, every choice). 1 min 33 s, 7 MB. Want to poke around inside? Try the brain scope.
Just the game: W1b's gauntlet, one brain, five lives, 1-1 to 1-4 in one run (deaths on 1-3 included), then 2-1. 1 min 33 s.

W1b's run

  • 1-1cleared, first try
  • 1-2cleared, first try
  • 1-3cleared on the fourth trydeaths at x 747, 1,200, 1,202
  • 1-4cleared, first try (Bowser)
  • 2-1last two lives lostx 343, 346; never trained on

Held-out clears, out of 50 starts each:

7/50
1-1
8/50
1-2
9/50
1-3
27/50
1-4

Last updated 2026-11-16 07:00 EST. Fabius and Cato2 trained on Cadence 0.95.0, the Scipio and Marcellus brains on 0.99.0. Older brains trained on Cadence (cadence-net) 0.91.1 and re-scored on 0.91.3: identical answers, identical scores. The 10/31 overnight brains trained on 0.91.3. Cadence is by Pragma Research; this experiment is independent.

The goal: World 8, never coached

Clear all four World 8 levels with at most 1 death in total

World 8 is the final world. No brain is ever coached on it or trained with a coach's labels there: whatever it learns on World 8 it learns by its own play. The gate is 8-1 to 8-4 cleared with at most one death across all four levels. The plan coaches worlds 1 to 7. (Until 11/6 the exam was World 4, and before that World 2.) Nothing on this page has played World 8 yet. So far Fabius, Cato2 and Scipio are coached on worlds 1 to 6 (24 levels), with World 7 held out as a development check.

The coach is verified on worlds 1 to 7: 164 of 168 test runs (6 starts per level) reach the flag, 11/8, assembled from six runs at the current settings (not one fresh pass). Worlds 1 to 3 72 of 72; worlds 4 to 7 92 of 96. Only 5-3 and 6-3 are short, 4 of 6 each (long leaps off high lifts). COACH.md. Planning lives in GitHub issues: #1 roadmap, #2 World 8, #4 plateau test, #11 Fabius vs Cato2, #12 Scipio, #13 Marcellus.

Scipio (11/14 to 11/15): does a slow memory fix the stalls?

No. The slow memory did not change the stalls (registered read at chunk 50). A likely reason, found after the read, is in the teaching.

Sean's question: is memory the answer to the holding problem? In the Fabius study every stuck attempt we checked (no new furthest x for 15 s; 108 attempts from Cato2 A and B at chunk 100 and Fabius A at chunk 65) held one button for its last 100+ decisions. On a pipe the screen looks the same on decision 5 and decision 150, so a brain that sees only the present can't tell it has been pushing against the same wall for seconds. Earlier memory tests could not catch this: Cicero's button memory faded within about 5 decisions, and the plateau test scored distance only.

All four: Cadence 0.99.0, Cato2's exact eyes (including the colour close-up placed from Mario's position in the game's memory), 256 neurons, 7 buttons, coached on worlds 1 to 6 with the coach labelling stalls up to 100 decisions, scored every 2nd chunk, stop at chunk 50. Registered before launch (11/14): the main number is stuck attempts per 192 practice attempts (8 starts on each of 24 levels), averaged over the last 10 scored chunks, with the 24-level mean x as a guard. The verdict also needs a time-only test on the same checkpoint: the memories cut to hold only the previous moment, so a change in stalls is pinned on how long the memory lasts, not on removing an input. ScipioA started 11/14 03:01, the others 05:44 to 05:46. Issue #12.

The read (11/16)

runstuck of 192never left the startdeaths of 192flags of 19224-level mean x
ScipioA (slow memory)44.414.8147.20.4807
ScipioB (slow memory)37.314.3153.41.3837
Twin (no memory, control)45.514.1145.01.5827
ScipioFast (fast memory, one seed)25.00.0166.20.8884

Means over the last 10 scored chunks (32 to 50), 8 practice starts on each of 24 levels. Registered rules: the memory brains had to beat Twin by more than 10 stuck attempts, and a time-only test (each memory cut to hold only the previous moment, same strength) had to raise their stuck count by more than 10. Neither happened: ScipioA sits 1.1 below Twin and ScipioB 8.2 below, and the time-only test changed stuck by -6.7 (A) and +9.3 (B), the mean over chunks 46, 48 and 50 (per chunk A -14, -16, +10; B +13, 0, +15). The test passed its own check (agreement with the coach fell 0.04, under the 0.05 limit), and the untouched brains replayed every one of their 192 attempts exactly as in training. Verdict: the memory was not used on stalls (this test cannot tell "not used" from "useless"). The time-only cut did cost distance (47 to 402 mean x by checkpoint), but switching the memories off entirely cost less on average (A 157 against 72, B 203 against 40), so that loss reads as a disturbance, not as the memory helping. ScipioFast's lead is mostly in starts it never stood at (4-2 and 4-4); under the time-only test its stuck count rose 13.7 (unregistered, one seed, no verdict).

What was actually holding them

These three were found after the registered read, so they likely explain it; they are not part of it. At the 4-4 training start the coach itself waits on about half the frames, so 4-4 is not a tie-rule stall. Details: issue #12.

Marcellus (since 11/16): clean labels on stalls

The next brain is Twin with one change: in a stall (10 decisions without progress) and before the first step forward, the coach's own label stands; the tie rule only applies while the brain is moving. Two seeds, each paired with a control that differs by that rule alone (MarcellusA starts from Twin's exact brain; MarcellusB and the new TwinB share seed 1). Registered before launch: stuck per 192 over the last 10 scored chunks at chunk 50; "labels help stalls" if both pairs beat their control by more than 10. My prediction, written down before launch: 4-2 stalls go to about 0, 2-2 drops but not to 0, and the main number passes but only just. Memory comes back after this, on the clean labels, with two seeds. Issue #13.

Public gauntlet (every 10 chunks)

Sean's idea: every 10 chunks each Scipio brain gets 10 lives in total to play the game from 1-1 in order. Greedy, no learning, a flag moves on to the next level, a death or 15 s without progress costs a life. It stops after 7-4, so World 8 stays unseen. This is for this page only and never decides anything; the registered numbers are in the Scipio section above. Different rules from the older five-life gauntlets further down.

chunkbrainfurthestlevels clearedlives usedclip
50 (11/15 10:49)ScipioA1-2, x 9731-110 of 10clip
50 (11/15 14:50)ScipioB1-2, x 9031-110 of 10-
50 (11/15 12:49)Twin1-1, x 901none10 of 10-
50 (11/15 09:39)ScipioFast1-1, x 898none10 of 10-
40 (11/15 09:21)ScipioB1-2, x 1,1901-110 of 10clip
40 (11/15 05:09)ScipioFast1-2, x 9451-110 of 10-
40 (11/15 05:31)ScipioA1-1, x 2,757none10 of 10-
40 (11/15 07:01)Twin1-1, x 898none10 of 10-
30 (11/14 22:39)ScipioFast1-1, x 723none10 of 10-
30 (11/14 22:10)ScipioA1-1, x 596none10 of 10clip
30 (11/15 01:59)ScipioB1-1, x 434none10 of 10-
30 (11/15 00:07)Twin1-1, x 315none10 of 10-
20 (11/14 18:07)ScipioFast1-1, x 2,386none10 of 10clip
20 (11/14 19:46)ScipioB1-1, x 1,789none10 of 10-
20 (11/14 16:36)ScipioA1-1, x 1,433none10 of 10-
20 (11/14 19:27)Twin1-1, x 899none10 of 10-
10 (11/14 11:46)ScipioA1-1, x 315none10 of 10clip
10 (11/14 11:47)ScipioB1-1, x 315none10 of 10-
10 (11/14 11:47)ScipioFast1-1, x 315none10 of 10-
10 (11/14 11:48)Twin1-1, x 315none10 of 10-
ScipioFast at chunk 20: ten lives on 1-1, the furthest of the four (x 2,386 of the level's roughly 3,160). 151 s. The chunk-10 clip (ScipioA) is linked in the table.

All four brains played the same way at chunk 10: every life ended at x 312 to 315 on 1-1, about two seconds in. Their practice scores on 1-1 say the same (mean x 314 on all 8 starts). Each chunk-10 checkpoint had about 5 to 6 hours of training. At chunk 20 (11/14 afternoon and evening) all four still spent their 10 lives on 1-1, but got further: furthest x 2,386 (ScipioFast), 1,789 (ScipioB), 1,433 (ScipioA), 899 (Twin). One gauntlet per brain is a public look, not a test; these numbers decide nothing.

Chunk 30 went backwards for the three memory brains: every one of their lives on 1-1 ended standing still (15 s without progress, x 434 to 723), and Twin died 10 times at x 315. At chunk 40 ScipioB and ScipioFast cleared 1-1 for the first time in a gauntlet and died on 1-2; ScipioA got to x 2,757 on 1-1 without the flag. At chunk 50 (the last checkpoint) ScipioA and ScipioB cleared 1-1 and died on 1-2; ScipioFast and Twin spent all 10 lives on 1-1. No brain has passed 1-2. One gauntlet per brain is a public look; it swings from round to round and decides nothing.

Fabius vs Cato2 (11/8 to 11/14): recall memory plus prediction

Short answer: Fabius does not beat Cato2. With two seeds each they can't be told apart.

Fabius A/B: 3,536 inputs (body sense, a grey view of the whole screen, a colour close-up around Mario; the close-up is placed using Mario's position read from the game's memory, but the brain gets only its pixels, never the position numbers), a recall memory and 104 prediction slots, Cadence 0.95.0. Cato2 A/B: the same eyes and the same teaching, with recall off and no prediction slots. Both coached on worlds 1 to 6 (24 levels), 15 s stuck limit. Registered before launch (11/8): one number per run, the 24-level mean x on 8 practice starts per level, averaged over chunks 81 to 100. Fabius wins only if it beats Cato2 by more than the larger gap between two seeds of one brain.

runmean x, chunks 81 to 100flags, of 3,840 attempts
Cato2 A (control)990.469
Cato2 B (control)925.684
Fabius A905.819
Fabius B935.740

Also not registered: both Cato2 seeds stand still on 4-4 at every chunk from 81 to 100. Deviation: the registered chunk-25 recall check (recall switched off against on) was not run on time; it ran late on 11/14 and passed (Fabius A on 817 / off 302, Fabius B on 638 / off 250, floor 242, the Cato2 seed gap on the same 6 levels at chunk 25). Switching recall off removes a whole input, so this shows the brains lean on that input, not that they use it as memory. The chunk-10 check passed on time (Fabius A on 374 / off 307, Fabius B on 772 / off 314, floor 288). Practice starts only; none of these brains has been scored on held-out starts. Fabius A and Cato2 A ran on, unregistered, to about chunks 112 and 116 (not read); then all four were stopped. Issue #11.

Plateau test (11/5 to 11/6): does memory plateau higher?

Short answer: no verdict either way. None of the three memory recipes plateaus higher, and none plateaus lower.

Sean's hypothesis: brains with and without memory learn at about the same pace, but memory plateaus higher. The six cold-start brains below (Cato, Cicero, Livy, two seeds each) resumed exactly where they stopped and trained on to chunk 200, joined by a new arm, CiceroLow (Cicero with its working memory turned down from 3.0 to 0.3, since Cadence's docs say the strong setting hurts imitation). All 8 runs reached chunk 200.

The test, as written before launch (11/5)

One number per run: the mean over the 12 training levels of the best x on 8 fixed practice starts, averaged over chunks 191 to 200. (This is the 8-start version of the curve in the cold-start section, so it isn't directly comparable to the AUC / END table there.) Noise floor = the biggest gap between the two seeds of one brain type. Memory "plateaus higher" only if BOTH seeds beat Cato's matching seed by more than the floor; "plateaus lower" by the mirror rule; anything else: not distinguishable.

runmean best x, chunks 191 to 200slope over the last 60 chunks (x per chunk)
Cato A (control)1,160-1.0 ± 0.8
Cato B (control)1,310+1.9 ± 1.1
Cicero A1,159+1.4 ± 1.0
Cicero B1,281+1.9 ± 0.9
Livy A1,251+2.5 ± 1.1
Livy B1,138+0.9 ± 0.9
CiceroLow A1,223-0.1 ± 0.9
CiceroLow B1,254+1.2 ± 0.8

Five of the six seed comparisons are inside the floor. With two seeds per brain this test only sees big effects; a small memory benefit or cost is not ruled out. 6 of 8 curves still slope up slightly at chunk 200 (about +1 to +2.5 x per chunk; only Cicero B and Livy A clearly above noise), so "plateau" is loose. Correction on record: an interim text on 11/6 said "Livy plateaus lower" from a chunk-161 read (floor 51 then); at chunk 200 it is not distinguishable. These runs kept the old stuck limit (2.7 s) by design, since the test was registered with it. Notebook: COLDSTART.md, issue #4.

Cold-start night (11/2 to 11/3): memory from birth, World 4 as the exam (then; since 11/6 the exam is World 8)

Short answer: no difference we can tell apart from luck. World 4 at first sight: 0 clears.

The memory night started from a brain that already played well without memory, so maybe it never needed one. This time six brains started from nothing: Cato (no memory), Cicero (thought + button memory) and Livy (all four memories), two seeds each, memory live from the first lesson. Same coach, same lessons, about 20 hours, trained on all 12 levels of worlds 1 to 3. World 4 was never coached or trained on: it was the exam then. (Since 11/6 the exam is World 8, and World 4 becomes a coached training world.) These six brains continued in the plateau test above.

The test, as written before the night

One number per brain: the practice-start mean x averaged over all 12 training levels, as a learning curve. Compared at equal lessons (chunks 0 to 74, 37,000 lesson batches, the most every run reached). AUC = the area under that curve divided by its length (a mean x over training); END = the value at chunk 74. "Noise" = the biggest gap between the two seeds of one brain type. Memory "helps" only if BOTH seeds beat Cato's matching seed by more than the noise, on BOTH numbers.

runAUC (mean x)END (mean x at chunk 74)practice flags at 74, of 288
Cato A (control)7839790
Cato B (control)8011,0918
Cicero A7911,0770
Cicero B73592712
Livy A7201,1190
Livy B7389660

World 4, first sight (not coached or trained at the time)

checkpoint12 training levels, held-out flags of 6004-1 to 4-3 flags of 1504-1 to 4-3 mean x
Cato A ck74 / ck1047 / 30 / 0520 / 513
Cato B ck74 / ck848 / 110 / 0602 / 751
Cicero A ck74 / ck1081 / 150 / 0571 / 414
Cicero B ck74 / ck9311 / 180 / 0412 / 595
Livy A ck74 / ck903 / 70 / 0528 / 387
Livy B ck74 (its last)60344

Greedy play on 50 held-out starts per level, checkpoints fixed in advance (chunk 74 = equal lessons, and each run's last). The few training-level clears are almost all 1-4, 2-4 and 3-2. 4-4 (the maze castle) is reported separately: every attempt of every brain ends at exactly x 138: the lava pit right after the start. We checked it's a real death, not a bug (a simple scripted jump pattern gets past it), so no brain ever jumps that first pit. Two seeds per brain, fresh brains, about 20 hours: the honest expectation was near-zero clears, and the curves are the result.

World 4 first sight: Cato B ck84 (greedy) from the start of 4-1, five lives. Deaths at x 346, 1,842, 1,842, 1,320, 344. Its best World 4 run of the night.
The only level cleared in 48 cold-start gauntlets: Cicero A ck108, sampled (temperature 0.05), clears 1-1 on its fourth life, then dies twice on 1-2. A demo, not evidence.

Memory night (11/1 to 11/2): does a memory help?

Short answer: not yet. Every pre-registered test lost.

Four brains trained side by side overnight from the same start (W1_F2ck110), same coach, same lessons, same driver, about 13 to 16 hours each. One has no memory (Cato, the control). The others got memory parts that Cadence already has but that our Mario play had never switched on. None of them beat the control by more than the gap between two copies of the control itself, and switching the short-term memory off at play made the best memory brain better, not worse.

How it remembers (plain words)

Every memory brain started as an exact copy of W1_F2ck110 with its memory wires at zero, so on the first chunk all four played identically (checked: same buttons, same scores). Anything different later came from what the memory wires learned.

Cato
the control: no memory
331-1
101-2
461-3
301-4

Top 3 picks: 119, 70, 33 of 200 (mean 74). Scores shown: its first pick, ck20.

memory parts: none · 15 h 47 min

AgrippaX
the control's twin
291-1
141-2
41-3
381-4

Has a place memory (keyed on x), but the night's recipe never writes to it, so it trained like Cato. Top 3: 85, 64, 60 (mean 70). It measures the noise.

memory parts: Place memory (empty) · 15 h 47 min

Cicero
thinks back a moment
191-1
111-2
131-3
451-4

Top 3: 88, 97, 61 (mean 82). The brain in the World 1 demo below. With its memory switched off it scores 106.

memory parts: Thought memory, Button memory · about 12 h 55 min, less after a CPU slot jam

Livy
all four memories
51-1
181-2
31-3
171-4

Top 3: 43, 65, 56 (mean 55), the lowest. Its weights lean on the memories (all off: 15), but it plays worse than the control.

memory parts: Thought memory, Button memory, Echo, Place memory · about 15 h, ~3x the work per lesson

Agrippa
remembers places across lives
501-1
111-2
61-3
251-4

W1_F2ck110's weights plus a place memory, tested only on retries (11/1 afternoon): no better than reloading the brain every life.

memory parts: Place memory (keyed on the screen) · not trained

Scores: solo clears of 50 held-out starts per level, World 1. Each brain's "top 3" are its three best checkpoints on practice starts (both legs of the night pooled, the shared starting checkpoint left out), then played once on the 50 held-out starts. Cards show each brain's first pick. W1_F2ck110, where they all started: 50 / 11 / 6 / 25 = 92.

The tests, as written before the night

brain (top 3 picks)World 1 clears, of 200mean2-1 first sight, mean x2-1 starts past the pipe (x 723), of 50memory off at play (first pick)
Cato (control)119, 70, 33746710, 0, 0-
AgrippaX (control's twin)85, 64, 60705890, 5*, 0no change (85)
Cicero88, 97, 61825240, 0, 7+18 (88 → 106)
Livy43, 65, 56555720, 0, 0-28 (43 → 15)

*AgrippaX's five (and five of W1_F2ck110's) are deaths at x 735, right at the pipe; Cicero's seven, all on one checkpoint, reach x 914 to 1,021, really past it. Nobody cleared 2-1 or 2-2. World 2 was not coached that night (it joined training on 11/2).

How big is the noise?

Cato and AgrippaX are two copies of the same recipe (AgrippaX's place memory stays empty all night, and its extra input never reaches the decisions), so the gap between them is luck. Their top-3 means differ by only 4 clears (74 vs 70), but single checkpoints from one line swing from 33 to 119 of 200, and the twins' best picks differ by 34. A rough estimate from that spread: a top-3 difference under about 25 clears is noise. Cicero (+8) and Livy (-19) are both inside it. On 2-1 the twins' means differ by 82 x. The 1-3 record (Cato_ck20, 46/50) is one lucky draw from many.

Also unequal: Cicero got about 3 hours less (it stopped itself after 3 hours without a new best and was restarted, then waited for a CPU slot), and Livy's lessons cost about 3x as much, so it saw fewer. One seed per brain. The next nights fixed most of this: equal lessons and a rule fixed in advance. The plan said 3 seeds per brain; both later tests (cold start, plateau) used 2.

Retry curves: does it learn from dying?

Livy (first pick, ck57) plays 10 lives in a row from each of 8 held-out starts. Remembers: its place notes are kept between lives. Control: the same brain reloaded before every life. Median furthest x per life; the weights never change.

remembers between livesreloaded every life (control)
Retry numbers as a table
level, armmedian furthest x, lives 1 to 10clears in 80 lives
1-2, remembers1646, 863, 870, 861, 879, 851, 69, 817, 446, 90410
1-2, control1646, 1531, 1670, 1198, 1392, 859, 929, 1049, 67, 132018
2-1, remembers322, 731, 417, 465, 436, 398, 626, 461, 450, 5010
2-1, control322, 399, 530, 348, 401, 530, 461, 498, 530, 5300

Lives 7 to 9 start after a long idle wait, so an enemy often reaches Mario right away: both arms dip there.

Why it fails, best guess: a death note only says "not the button I pressed here", never what to do instead. With 7 buttons and a brain that already likes run and run+jump, the notes mostly push it onto a worse second choice.

Demo: a memory brain beats World 1 (one run, not evidence)

Cicero (thought + button memory, checkpoint r2 ck1) at 01:03: 1-1 first try, 1-2 and 1-3 on the second try, 1-4 first try, then three lives on 2-1. 1 min 41 s.

The second brain ever to beat all of World 1 in one five-life run, and the first with a memory. It is a demo: one run of one checkpoint, played with a little randomness in the button choice (sampled at temperature 0.05; the held-out scores are plain greedy play). The same brain clears 19 / 11 / 13 / 45 of 50 held-out starts, and with its memory switched off it clears more (26 / 11 / 24 / 45). So this clip shows a capable brain, not a memory that helps.

Cicero's brain-cam, 1-2 from held-out start 38 (a clear). Same panels as W1b's, plus its memories live: the thought memory (256 prefrontal neurons carrying the last moments) and the button memory (what it pressed, fading). Greedy play, recorded during the run; the recorded choice matched the pressed button at every decision. 21 s. Also: 1-1 from the real start.

Best per level

50/50
1-1 · W1_F2ck110
29/50
1-2 · 1-2_12n59
46/50
1-3 · Cato_ck20*
50/50
1-4 Bowser · U3ck28

Solo clears on 50 held-out starts, each level's brain from the scoreboard below. These are four different brains; the one-brain result is W1b above. *Best of many: the 11/1 night saved about 350 checkpoints across four brains and this is the best 1-3 among the twelve scored on held-out starts. A record pick, not a typical brain (the same line's other picks score 11 and 12).

One brain, every World 1 level

W1_F2ck20 clears 1-1 from held-out start 34. 22 s.
Same weights, 1-2 (underground) from held-out start 34. 22 s.
Same weights, 1-3 from held-out start 99, its only 1-3 clear of 50. 16 s.
Same weights, 1-4 and Bowser from held-out start 34. 13 s.

How it got there (overnight 10/31): the 1-2 brain line (1-2_12n59) practised 1-2 for 3 hours, then 1-3 and 1-4 were added back with lessons from all four levels and coaching on the weak level, picking checkpoints on practice starts only. Checkpoint 20 of that run (W1_F2ck20, about 45 minutes in) was the first whose practice starts flagged all four levels; the 50 held-out starts above were played after it was picked. In the five-life gauntlet W1_F2ck20 lost all five lives on 1-1 (it clears 1-1 only about a third of the time).

One brain, three levels

U3ck83 clears 1-1 from held-out start 33. 22 s.
Same brain, same weights: 1-3 (tree tops, moving lifts) from held-out start 33. 15 s.
Same brain again: 1-4, firebars and Bowser, from held-out start 33. 13 s.

Its sibling U3ck44 ("the balanced one") also clears all three: 25 / 15 / 37 of 50. Both were tested on 1-2 on 10/31 (74 starts each): 0 clears, both stop around x 900.

Scoreboard

levelsolo clearsbrainstatus
1-150/50W1_F2ck110perfect on held-out starts, and the same brain clears all four World 1 levels
1-229/501-2_12n59not yet: up from 8/50 overnight; clears more than half the time
1-346/50Cato_ck20 (best of many)consistent; the same brain also clears 1-1 33, 1-2 10, 1-4 30 (119/200). Before: one_11_13 42/50, U3ck83 45/50
1-4 Bowser50/50U3ck28perfect on held-out starts; the same brain also clears 1-1 on 30/50
2-10/50cold-start brains (best of 11 checkpoints)a training level since 11/2, still never cleared: best mean x 931. At first sight (one_11_13, 10/31) it reached x ~330 (random buttons ~380)
2-20/50cold-start brains (best of 11 checkpoints)a training level since 11/2, still never cleared: best mean x 1,266. At first sight (one_11_13, 10/31) it reached x ~210 (random buttons ~1,230)
2-3 to 3-4best 10/50cold-start brains (best of 11 checkpoints)training levels since 11/2: 2-3 1, 2-4 9 (Cicero B ck93), 3-1 0, 3-2 10 (Cicero B ck74), 3-3 1, 3-4 0, of 50 each
4-1 to 4-30/150cold-start brains, first sightfirst sight before any coaching (11/3); World 4 is now a coached training world: best mean x 751 (Cato B ck84), furthest 1,859
World 8-none yetthe exam: never coached, own play only; gate = all four levels with at most 1 death in total

Solo clears out of 50 held-out starts. A start waits a different number of frames before the first button, which moves the enemies and lifts; none of these 50 were used to train the brain or to pick the checkpoint. "Consistent" = at least 40 of 50. Brains by name: see "Meet the brains". The best single brain across three levels is U3ck83 (15 / 45 / 33 on 1-1 / 1-3 / 1-4).

Meet the brains

Every brain is one Cadence region of 256 neurons reading the screen (coarse grey blocks, this frame and the last) and its own body (speed, ground/air, held buttons). All came from the same line of training (coach labels + DAgger, LEARNING_FIX.md), older ones trained on cadence-net 0.91.1 and re-scored on 0.91.3, W1 and F30 trained on 0.91.3. Scores: solo clears of 50 held-out starts.

W1_F2ck35 (W1b)
the balanced finisher
71-1
81-2
91-3
271-4

Beat all of World 1 in one five-life gauntlet; every level at least 7/50.

all four levels, run p4/F2 checkpoint 35 (0.91.3)

W1_F2ck110
the finisher with a perfect 1-1
501-1
111-2
61-3
251-4

Highest four-level total (92/200); perfect 1-1.

all four levels, run p4/F2 checkpoint 110 (0.91.3)

W1_F2ck87
the finisher with the best 1-3
221-1
71-2
151-3
271-4

Every level at least 7/50, 1-3 15/50.

all four levels, run p4/F2 checkpoint 87 (0.91.3)

W1_F2ck40
the finisher, best total
291-1
81-2
51-3
341-4

All four levels, 76/200.

all four levels, run p4/F2 checkpoint 40 (0.91.3)

W1_F2ck39
the finisher with a strong 1-1
431-1
11-2
41-3
361-4

All four levels, consistent on 1-1.

all four levels, run p4/F2 checkpoint 39 (0.91.3)

W1_F2ck20
the World 1 finisher
161-1
101-2
11-3
341-4

First brain to clear every World 1 level.

all four levels, run p4/F2 checkpoint 20 (0.91.3)

F30
the 1-2 + Bowser one
421-1
91-2
01-3
341-4

First brain to clear 1-2 and 1-4 together.

all four levels, run p4/F checkpoint 30 (0.91.3)

U3ck83
the three-level champion
151-1
01-2
451-3
331-4

Best three-level total (93/150), best 1-3.

1-1 + 1-3 + 1-4, run st_U3 checkpoint 83

U3ck44
the balanced one
251-1
01-2
151-3
371-4

First brain to clear all three practised levels.

1-1 + 1-3 + 1-4, st_U3 checkpoint 44

one_11_13
the two-level all-rounder
401-1
01-2
421-3
01-4

Consistent (40+) on two levels at once.

1-1 + 1-3, st_13d checkpoint 13

1-1_U3ck3
the 1-1 specialist
461-1
01-2
41-3
81-4

Best 1-1 of its generation (W1_F2ck110 later got 50/50).

1-1 + 1-3 + 1-4, st_U3 checkpoint 3

U3ck28
the Bowser specialist
301-1
01-2
111-3
501-4

Perfect 1-4.

1-1 + 1-3 + 1-4, st_U3 checkpoint 28

1-2_12n59
the underground one
141-1
291-2
01-3
01-4

Best 1-2 (29/50). It was the only brain that cleared 1-2 until the World 1 finishers (W1_F2*).

1-2 (90%) + 1-1, st_12n checkpoint 59

All brains as one table (easier to compare on a big screen)
brainnickname1-11-21-31-4trained on / checkpointwhat's special
W1_F2ck35 (W1b)the balanced finisher78927all four levels, run p4/F2 checkpoint 35 (0.91.3)beat all of World 1 in one five-life gauntlet; every level at least 7/50
W1_F2ck110the finisher with a perfect 1-15011625all four levels, run p4/F2 checkpoint 110 (0.91.3)highest four-level total (92/200); perfect 1-1
W1_F2ck87the finisher with the best 1-32271527all four levels, run p4/F2 checkpoint 87 (0.91.3)every level at least 7/50, 1-3 15/50
W1_F2ck40the finisher, best total298534all four levels, run p4/F2 checkpoint 40 (0.91.3)all four levels, 76/200
W1_F2ck39the finisher with a strong 1-1431436all four levels, run p4/F2 checkpoint 39 (0.91.3)all four levels, consistent on 1-1
W1_F2ck20the World 1 finisher1610134all four levels, run p4/F2 checkpoint 20 (0.91.3)first brain to clear every World 1 level
F30the 1-2 + Bowser one429034all four levels, run p4/F checkpoint 30 (0.91.3)first brain to clear 1-2 and 1-4 together
U3ck83the three-level champion15045331-1 + 1-3 + 1-4, run st_U3 checkpoint 83best three-level total (93/150), best 1-3
U3ck44the balanced one25015371-1 + 1-3 + 1-4, st_U3 checkpoint 44first brain to clear all three practised levels
one_11_13the two-level all-rounder4004201-1 + 1-3, st_13d checkpoint 13consistent (40+) on two levels at once
1-1_U3ck3the 1-1 specialist460481-1 + 1-3 + 1-4, st_U3 checkpoint 3best 1-1 of its generation (W1_F2ck110 later got 50/50)
U3ck28the Bowser specialist30011501-1 + 1-3 + 1-4, st_U3 checkpoint 28perfect 1-4
1-2_12n59the underground one1429001-2 (90%) + 1-1, st_12n checkpoint 59best 1-2 (29/50); the only 1-2 brain until the W1_F2 finishers

Newer brains (from 11/8), trained from scratch on worlds 1 to 6

These see more than the cards above: 3,536 inputs, the coarse grey screen plus a colour close-up around Mario. None has been scored on held-out starts yet, so they have no cards; their numbers are practice-start numbers in their own sections.

Gauntlet

Real start of 1-1, five lives. A flag moves on to the next level (1-2, 1-3, 1-4, 2-1 ...); a death restarts the current level; no learning between lives; game over on the fifth death. Retries start a few frames later (the emulator is deterministic, so an identical restart would replay the identical death). "One brain" = the same weights the whole way. Each level starts with small Mario. The newer 10-life gauntlets for the Scipio brains are in Public gauntlet.

timebraincleareddeaths (level: x)lives leftclip
11/03 08:55 ESTCicero A ck108 (cold start, from scratch; one brain, sampled play*)1-11-1: 679, 1,797, 1,129; 1-2: 918, 8860clip
11/03 08:53 ESTCold start, all six brains, greedy and sampled, from 1-1, 2-1, 3-1 and 4-1 (48 runs)none (except the row above)every run ends on its first level; best World 4: Cato B ck84 4-1: 346, 1,842, 1,842, 1,320, 34404-1 clip
11/02 07:05 ESTAgrippaX ck18 (memory night, first pick; one brain, sampled play*)1-11-2: 873, 873, 850, 871, 1,5950clip
11/02 06:44 ESTLivy ck57 (memory night, first pick; one brain, sampled play*)1-11-1: 315, 312, 312; 1-2: 1,743, 8670clip
11/02 01:03 ESTCicero r2 ck1, thought + button memory (one brain, sampled play*)1-1, 1-2, 1-3, 1-41-2: 1,344; 1-3: 936; 2-1: 471, 722 (stuck), 3300clip
11/01 23:47 ESTCato ck20, the no-memory control (one brain, sampled play*)1-1, 1-21-2: 978 (stuck); 1-3: 1,256, 1,140, 1,942, 1,9100clip
11/01 05:11 ESTW1_F2ck110 (one brain; held-out 50 / 11 / 6 / 25)1-11-2: 872, 883, 1,320, 468, 1,3210clip
11/01 04:42 ESTW1_F2ck87 (one brain; held-out 22 / 7 / 15 / 27)1-11-1: 2,471, 3,154 (stuck); 1-2: 1,744, 1,646, 1,7420clip
11/01 03:02 ESTW1_F2ck40 (one brain; held-out 29 / 8 / 5 / 34)1-11-1: 1,153, 1,153; 1-2: 851, 1,325, 1,3210clip
11/01 02:30 ESTW1b = W1_F2ck35, the balanced finisher (one brain)1-1, 1-2, 1-3, 1-41-3: 747, 1,200, 1,202; 2-1: 343, 3460clip, with brain
11/01 02:30 ESTW1_F2ck39 (one brain)1-11-1: 1,816, 2,469; 1-2: 955, 882, 6790clip
11/01 01:51 ESTW1_F2ck20, the World 1 finisher (one brain)none1-1: 2,850, 2,759, 2,754, 1,796, 2,0250clip
11/01 01:02 ESTF30 (one brain; clears 1-1, 1-2, 1-4 on held-out starts)1-1, 1-21-1: 1,790; 1-2: 1,744, 851; 1-3: 427, 4270clip
11/01 00:51death-clinic brain (phase 2, checkpoint 107) (one brain)none1-1: 2,471, 1,674, 1,674, 695, 6980clip
10/31 22:511-2 practice brain (phase 2, plain coaching, checkpoint 35) (one brain)1-1, 1-21-1: 1,790, 1,789; 1-2: 1,596; 1-3: 656, 6590clip
10/31 22:501-2_12n59, the underground one (one brain)1-1, 1-21-2: 1,595, 852, 1,336; 1-3: 700, 7000clip
10/31 21:23RAM-map brain, end of phase 1 RAM / cheat1-11-2: 1,651, 291 (stuck), 676, 673, 8510clip
10/31 19:20fresh brain, normal senses (1 h 30 min old, phase 1 control)none1-1: 1,126 x50clip
10/31 19:20fresh brain reading a RAM map RAM / cheat (1 h 30 min old)none1-1: 2,850 (stuck), 1,820, 1,820, 2,850 (stuck), 1,8200clip
10/31 17:48relay of specialists: 1-1_U3ck3, 1-2_12n59, U3ck83 (1-3), U3ck28 (1-4), U3ck83 (2-1 on)1-1, 1-2, 1-3, 1-41-2: 1,595, 852, 1,336; 2-1: 531 (stuck), 723 (stuck)0clip
10/31 17:48U3ck44 (one brain)1-11-1: 2,762, 1,784; 1-2: 323, 354, 354 (all stuck)0clip
10/31 17:11U3ck83 (one brain)1-11-2: 322, 470, 322, 893, 3540clip

Best single brains: W1b (11/01 02:30) and Cicero (11/02 01:03), all four World 1 levels in one run each. Gauntlets are demos: one run of one checkpoint (W1b was one of six checkpoints tried that night). *The memory-night gauntlets and the "sampled" cold-start runs choose buttons with a little randomness (sampled at temperature 0.05, weights unchanged, checked); the older rows are greedy. Found and labelled 11/02.

Not one brain: a relay of specialists (four brains, one per level) clears all of World 1 in one five-life run, then loses two lives on 2-1, which none of them ever trained on. 1 min 53 s, deaths included.

Tests from the 10/31 night

Would perfect "what is where" help? RAM / cheat

A ceiling test, labeled as a cheat: a fresh brain read a map built from the game's memory (13 x 16 tiles around Mario: solid, enemy, pit, power-up) plus its body sense, no screen. A matched fresh brain read the normal screen. Same coach, same recipe, same 3.5 hours, same share of the PC.

fresh brain, 3.5 h1-11-21-31-42-1 first sight
RAM map cheat, best 1-4 checkpoint00049/50x 1,021*
RAM map cheat, best joint checkpoint10/500036/50x 429
normal screen (control), best checkpoint2/50000x 340-530

Result: with a perfect map it learns World 1 levels far faster (5 more hours: 1-1 50/50, 1-4 43/50), but it never cracked 1-2 or carry over to World 2 in 3.5 h, which were the pre-registered win conditions. So by its own rules the test lost. *Not pre-registered and the best of five brains scored, so treat it as a lead, not a result. This brain never plays on this page's scoreboard; it can't see the screen.

A "death clinic" for 1-2

When a solo run dies, rewind to 1, 2 and 4 seconds before the death, let the brain try again from there with the coach labeling, and keep its own moves when it gets past the spot where it died. Compared against plain coaching with the same coaching budget, both starting from the 1-2 brain (29/50).

3 hours of 1-2 practice1-2 clearswrong on must-get-right moments, along its own path
start (1-2_12n59)29/506.7%
death clinic18/5016.0%
plain coaching17/506.5%

Result: the clinic worked mechanically (about half its retries got past the old death spot) but didn't help. Neither arm beat the starting brain in 3 hours. One run each, so a small effect could hide in the noise; a big one would have shown.

Latest

Older entries (10/29 to 11/01 01:49)
  • Milestone: one brain (W1_F2ck20) clears all four World 1 levels with the same weights on held-out starts: 1-1 16/50, 1-2 10/50, 1-3 1/50, 1-4 34/50. Clips above. Not consistent on 1-3 or 1-2 yet.
  • Death clinic for 1-2 (practice from just before each death): no gain over plain coaching, and neither beat the 29/50 starting brain in 3 hours. Now trying to grow one brain that clears all four World 1 levels.
  • New brain F30: 1-1 42/50, 1-2 9/50 and 1-4 34/50 with the same weights, the first brain to clear both 1-2 and 1-4. 1-3 is still missing; training continues.
  • Gauntlet: one brain (1-2_12n59) clears 1-1 and 1-2 in a single five-life run, the furthest a single brain has gone, then loses its last two lives on 1-3.
  • RAM-map ceiling test cheat: a fresh brain with a perfect "what is where" map learns 1-4 to 49/50 in 3.5 h, where the same fresh brain on the screen gets almost nothing. But no 1-2 and no World 2, so it misses its own bar. Next: "death clinic" on 1-2 (practice the exact spots where it dies).
  • Gauntlet: a relay of specialists (one brain per level) clears all four World 1 levels in one five-life run (1-2 on the fourth try), then dies on 2-1. Single brains stop at 1-2.
  • Re-scored every brain on Cadence 0.91.3: identical. Tested the two three-level brains on 1-2: 0 clears in 74 starts each, so "every World 1 level" is not claimed. New: brain roster and a 5-life gauntlet (first one: U3ck83 clears 1-1, then loses all five lives on 1-2).
  • First try at World 2 with no coach and no new learning: 0 clears for all five best brains. 2-1: furthest x 723 (stuck at the first tall pipe), most die within 2 seconds. 2-2 (underwater, never seen water): furthest x 514, mostly stops pressing buttons. World 1 skills don't transfer yet. Clips: 2-1, 2-2.
  • 1-2 up again: 29/50 held-out starts (1-2_12n59).
  • 1-2 moves: 23/50 held-out starts (was 8/50), after testing many more checkpoints between coaching rounds.
  • New best on 1-3: 45/50 (U3ck83). That checkpoint also clears 1-4 33/50 and 1-1 15/50.
  • First checkpoint that clears all three practised levels at least sometimes (U3ck44): 1-1 25/50, 1-3 15/50, 1-4 37/50 (not consistent yet).
  • 1-4 on 50 of 50 held-out starts (checkpoint D, now named U3ck28, which also clears 1-1 on 30/50 and 1-3 on 11/50).
  • 1-4 goes consistent: 42/50 held-out starts, Bowser beaten (checkpoint C, 1-4 only).
  • New best on 1-1 alone: 46/50 held-out starts (a checkpoint that does not hold 1-3).
  • One brain (A, now named one_11_13) holds two levels: 1-1 40/50 and 1-3 42/50.
  • 1-4 beaten solo, Bowser included, on 38/50 held-out starts (checkpoint B, joint_11_14_ck10).
  • Correction: a 19:31 claim that a brain cleared 1-3 on all 17 practice starts rested on starts of one frame phase. Fixed the test; brain A is measured the fixed way.
  • 1-1 on 44/50 held-out starts (joint_11_13_ck8).
  • 1-2 parked after ~5 hours: best 8/50.
  • First 1-1 brain to clear 10 of 10 test starts (38/50 on the wider test).
  • First ever solo clear of 1-1 by a Cadence brain.

Clips

one_11_13 clears 1-1 from a held-out start. 22 s.
The same brain, same weights, clears 1-3 (tree tops and moving lifts) from a held-out start. 16 s.
U3ck28, the Bowser specialist (50/50 on 1-4), beats 1-4 from a held-out start: firebars, then over Bowser to the axe.
An earlier 1-4 brain (joint_11_14_ck10, 38/50) beats 1-4 from a held-out start: firebars, then a jump over Bowser to the axe. 14 s.
joint_11_13_ck8 (44/50 on 1-1) on a held-out start. 20 s.

Every button in every clip is the brain's.

How it learned

Honest caveats

More Mario

Brain scope

Scrub through a solo run and watch the brain's real neuron activity, or poke it to see what part of the screen each move depended on.

Level by level

The in-order run, first-sight tests, per-level tables and more clips.

The first 1-1 run

10/28-29: the older brain's replay, learning curve and rescued runs.

Watch the first run

The same older run as plain video: alone, and with coach rescues.