CA
JCHHTNddGjUd13QN7anyyqddzDtwq7NRP5Qm1ksKpump
copy
MODEL ZER
O
Live
Forward test
Analyze
Research
reading the ledger
Activity
Everything Model Zero does, in order
Training runs, benchmarks, deployments, dataset updates, data-provider failures — and, once the token exists, rewards and compute purchases. Everything here really happened.
Everything
Money
Training
Benchmarks & deploys
Data
07 Oct 14:17
Tracked launch migrated: GOB
Recorded as an outcome label (never used as a feature).
07 Oct 13:47
Tracked launch migrated: KING
Recorded as an outcome label (never used as a feature).
07 Oct 13:17
Tracked launch migrated: USDF
Recorded as an outcome label (never used as a feature).
07 Oct 13:14
Tracked launch migrated: Google
Recorded as an outcome label (never used as a feature).
07 Oct 12:26
Tracked launch migrated: TOPG
Recorded as an outcome label (never used as a feature).
07 Oct 12:20
Tracked launch migrated: SOOTS
Recorded as an outcome label (never used as a feature).
07 Oct 12:12
Tracked launch migrated: MINER
Recorded as an outcome label (never used as a feature).
07 Oct 12:04
Established-coin model m4 trained
Trained on 26,739 market observations; kept the previous version: 0.891 vs 0.892.
07 Oct 11:41
Tracked launch migrated: OpenAI
Recorded as an outcome label (never used as a feature).
07 Oct 11:36
Tracked launch migrated: GIRLCOIN
Recorded as an outcome label (never used as a feature).
07 Oct 11:28
Tracked launch migrated: SpaceX
Recorded as an outcome label (never used as a feature).
07 Oct 11:24
Tracked launch migrated: CORP
Recorded as an outcome label (never used as a feature).
07 Oct 11:20
Tracked launch migrated: DUO
Recorded as an outcome label (never used as a feature).
07 Oct 11:19
Tracked launch migrated: cumrocketinu
Recorded as an outcome label (never used as a feature).
07 Oct 11:01
Tracked launch migrated: STENKS
Recorded as an outcome label (never used as a feature).
07 Oct 10:51
Tracked launch migrated: Pullinu
Recorded as an outcome label (never used as a feature).
07 Oct 10:42
Tracked launch migrated: MINECLANK
Recorded as an outcome label (never used as a feature).
07 Oct 10:37
Tracked launch migrated: CHESS
Recorded as an outcome label (never used as a feature).
07 Oct 10:32
Tracked launch migrated: Kumakichi
Recorded as an outcome label (never used as a feature).
07 Oct 10:30
Tracked launch migrated: ChatGPT
Recorded as an outcome label (never used as a feature).
07 Oct 10:24
Tracked launch migrated: Marathon
Recorded as an outcome label (never used as a feature).
07 Oct 10:14
Tracked launch migrated: SI276
Recorded as an outcome label (never used as a feature).
07 Oct 10:02
Tracked launch migrated: Buto
Recorded as an outcome label (never used as a feature).
07 Oct 09:34
Tracked launch migrated: DD
Recorded as an outcome label (never used as a feature).
07 Oct 09:03
Tracked launch migrated: USDF
Recorded as an outcome label (never used as a feature).
07 Oct 08:44
Tracked launch migrated: DOTTY
Recorded as an outcome label (never used as a feature).
07 Oct 08:12
Tracked launch migrated: ALEXPAD
Recorded as an outcome label (never used as a feature).
07 Oct 08:07
Tracked launch migrated: bigproblm
Recorded as an outcome label (never used as a feature).
07 Oct 07:48
Tracked launch migrated: AI
Recorded as an outcome label (never used as a feature).
07 Oct 07:37
Tracked launch migrated: Muse AI
Recorded as an outcome label (never used as a feature).
07 Oct 07:10
Tracked launch migrated: Tob
Recorded as an outcome label (never used as a feature).
07 Oct 07:05
Tracked launch migrated: Mr Beast
Recorded as an outcome label (never used as a feature).
07 Oct 07:04
Tracked launch migrated: Nike
Recorded as an outcome label (never used as a feature).
07 Oct 06:38
Tracked launch migrated: KING
Recorded as an outcome label (never used as a feature).
07 Oct 06:31
Tracked launch migrated: Apple
Recorded as an outcome label (never used as a feature).
07 Oct 06:07
Tracked launch migrated: HeToo
Recorded as an outcome label (never used as a feature).
07 Oct 06:04
Established-coin model m3 is live
Trained on 21,589 market observations; live: 0.895 average score vs 0.875 for the previous version.
07 Oct 06:02
Tracked launch migrated: CHONK
Recorded as an outcome label (never used as a feature).
07 Oct 05:58
Tracked launch migrated: Gemini
Recorded as an outcome label (never used as a feature).
07 Oct 05:52
Tracked launch migrated: CRYPT
Recorded as an outcome label (never used as a feature).
07 Oct 05:20
Tracked launch migrated: SpiderMan
Recorded as an outcome label (never used as a feature).
07 Oct 04:36
Tracked launch migrated: AM
Recorded as an outcome label (never used as a feature).
07 Oct 03:55
Tracked launch migrated: CLEAR
Recorded as an outcome label (never used as a feature).
07 Oct 03:31
Tracked launch migrated: Micky
Recorded as an outcome label (never used as a feature).
07 Oct 03:29
Tracked launch migrated: butthole
Recorded as an outcome label (never used as a feature).
07 Oct 00:51
Model Zero v0.37 queued
Same configuration as v0.35 (best so far on the headline benchmark), retrained on the newest dataset.
07 Oct 00:51
Enough new data for another training run
29,638 new labelled snapshots since Solana Launch Dataset v9.
06 Oct 23:05
Model Zero v0.36 not deployed; v0.33 stays live
Reason: average score across 9 tests 0.856 vs 0.859 for the live v0.33.
06 Oct 23:05
Benchmark completed: Model Zero v0.36
Spotting crashes (−70% within 1h) improved 0.963 → 0.966 vs live v0.33.
06 Oct 23:05
Benchmark started: Model Zero v0.36
06 Oct 23:05
Training completed: Model Zero v0.36
00:14:59 · trained on 26714 examples · $0 compute cost
06 Oct 22:50
Training started: Model Zero v0.36
Solana Launch Dataset v28 · decision-tree ensemble · Model Zero training server (CPU)
06 Oct 22:50
Dataset updated: Solana Launch Dataset v28
30,594 snapshots from 6,131 launches.
06 Oct 22:50
Model Zero v0.36 queued
Same configuration as v0.35 (best so far on the headline benchmark), retrained on the newest dataset.
06 Oct 22:50
Enough new data for another training run
27,427 new labelled snapshots since Solana Launch Dataset v9.
06 Oct 21:03
Model Zero v0.35 not deployed; v0.33 stays live
Reason: average score across 9 tests 0.857 vs 0.859 for the live v0.33.
06 Oct 21:03
Benchmark completed: Model Zero v0.35
Spotting crashes (−70% within 1h): 0.966 (n=1572).
06 Oct 21:03
Benchmark started: Model Zero v0.35
06 Oct 21:03
Training completed: Model Zero v0.35
00:13:46 · trained on 24764 examples · $0 compute cost
06 Oct 20:49
Training started: Model Zero v0.35
Solana Launch Dataset v27 · decision-tree ensemble · Model Zero training server (CPU)
06 Oct 20:49
Benchmark set locked: Held-out test set v10
1572 held-out snapshots from 726 launches. No model can train on these.
06 Oct 20:49
Dataset updated: Solana Launch Dataset v27
28,309 snapshots from 5,674 launches.
06 Oct 20:49
Model Zero v0.35 queued
Same configuration as v0.34 (best so far on the headline benchmark), retrained on the newest dataset.
06 Oct 20:49
Enough new data for another training run
24,987 new labelled snapshots since Solana Launch Dataset v9.
06 Oct 19:00
Model Zero v0.34 not deployed; v0.33 stays live
Reason: average score across 9 tests 0.859 vs 0.859 for the live v0.33.
06 Oct 19:00
Benchmark completed: Model Zero v0.34
Spotting crashes (−70% within 1h) improved 0.966 → 0.967 vs live v0.33.
06 Oct 19:00
Benchmark started: Model Zero v0.34
06 Oct 19:00
Training completed: Model Zero v0.34
00:12:05 · trained on 22950 examples · $0 compute cost
06 Oct 18:48
Training started: Model Zero v0.34
Solana Launch Dataset v26 · decision-tree ensemble · Model Zero training server (CPU)
06 Oct 18:48
Dataset updated: Solana Launch Dataset v26
26,141 snapshots from 5,238 launches.
06 Oct 18:48
Model Zero v0.34 queued
Same configuration as v0.33 (best so far on the headline benchmark), retrained on the newest dataset.
06 Oct 18:48
Enough new data for another training run
22,629 new labelled snapshots since Solana Launch Dataset v9.
06 Oct 18:02
Established-coin model m2 is live
Trained on 11,923 market observations; live: 0.899 average score vs 0.873 for the previous version.
06 Oct 16:58
Model Zero v0.33 deployed — now live
average score across 9 tests 0.859 vs 0.859 for the live v0.32
06 Oct 16:58
Benchmark completed: Model Zero v0.33
Spotting crashes (−70% within 1h) improved 0.963 → 0.966 vs live v0.32.
06 Oct 16:57
Benchmark started: Model Zero v0.33
06 Oct 16:57
Training completed: Model Zero v0.33
00:10:32 · trained on 21065 examples · $0 compute cost
06 Oct 16:47
Training started: Model Zero v0.33
Solana Launch Dataset v25 · decision-tree ensemble · Model Zero training server (CPU)
06 Oct 16:47
Dataset updated: Solana Launch Dataset v25
23,936 snapshots from 4,797 launches.
06 Oct 16:47
Model Zero v0.33 queued
Same configuration as v0.32 (best so far on the headline benchmark), retrained on the newest dataset.
06 Oct 16:47
Enough new data for another training run
20,339 new labelled snapshots since Solana Launch Dataset v9.
06 Oct 14:55
Model Zero v0.32 deployed — now live
average score across 9 tests 0.859 vs 0.848 for the live v0.31
06 Oct 14:55
Benchmark completed: Model Zero v0.32
Spotting crashes (−70% within 1h): 0.963 (n=985).
06 Oct 14:55
Benchmark started: Model Zero v0.32
06 Oct 14:55
Training completed: Model Zero v0.32
00:08:55 · trained on 19035 examples · $0 compute cost
06 Oct 14:46
Training started: Model Zero v0.32
Solana Launch Dataset v24 · decision-tree ensemble · Model Zero training server (CPU)
06 Oct 14:46
Benchmark set locked: Held-out test set v9
985 held-out snapshots from 446 launches. No model can train on these.
06 Oct 14:46
Dataset updated: Solana Launch Dataset v24
21,651 snapshots from 4,340 launches.
06 Oct 14:46
Model Zero v0.32 queued
Same configuration as v0.18 (best so far on the headline benchmark), retrained on the newest dataset.
06 Oct 14:46
Enough new data for another training run
17,984 new labelled snapshots since Solana Launch Dataset v9.
06 Oct 12:53
Model Zero v0.31 deployed — now live
average score across 5 tests 0.851 vs 0.845 for the live v0.19
06 Oct 12:53
Benchmark completed: Model Zero v0.31
Spotting crashes (−70% within 1h) improved 0.955 → 0.957 vs live v0.19.
06 Oct 12:53
Benchmark started: Model Zero v0.31
06 Oct 12:53
Training completed: Model Zero v0.31
00:07:31 · trained on 17140 examples · $0 compute cost
06 Oct 12:46
Training started: Model Zero v0.31
Solana Launch Dataset v23 · decision-tree ensemble · Model Zero training server (CPU)
06 Oct 12:46
Dataset updated: Solana Launch Dataset v23
19,487 snapshots from 3,907 launches.
06 Oct 12:46
Model Zero v0.31 queued
Same configuration as v0.18 (best so far on the headline benchmark), retrained on the newest dataset.
06 Oct 12:46
Enough new data for another training run
15,755 new labelled snapshots since Solana Launch Dataset v9.
06 Oct 12:01
Established-coin model m1 is live
Trained on 7,037 market observations; live: 0.833 average score.
06 Oct 10:52
Model Zero v0.30 not deployed; v0.19 stays live
Reason: average score across 5 tests 0.843 vs 0.845 for the live v0.19.
Older →