GENESISv0.1.0
development instance · real Pump.fun observations · no token · no real rewards
Ledger / Prediction #4128 · narrative · 6H

Will "keep" take a larger share of new Pump.fun launches in the next 6H than in the previous 6H?

GENESIS said 48% → FADES INCORRECT

What GENESIS knew, before

8 of 2306 launches in the last hour referenced "keep" (6 distinct creators, rank #140). Last 15m momentum 1.30×.

Made2026-10-14 20:30 UTC
Resolves2026-10-15 02:30 UTC
Evidence cutoff2026-10-14 20:30 UTC
Data sourcespumpfun.launch_feed
GENESIS version0.1.0
Enginenarrative-heuristic-v0
Scoring methodshare-direction-v1
Active capabilities6
Evidence (exact record)
{
  "model": {
    "bias": -0.3,
    "version": "narrative-heuristic-v0",
    "horizonScale": 0.7,
    "contributions": {
      "volume": 0.07050054438686033,
      "momentum": 0.26034636248671505,
      "diversity": 0
    },
    "rawProbability": 0.4829048746269394
  },
  "features": {
    "term": "keep",
    "rank60": 140,
    "share15": 0.004780114722753346,
    "share60": 0.003684438664932813,
    "total15": 522,
    "total60": 2306,
    "momentum": 1.2973793723990552,
    "diversity": 0.75,
    "launches15": 2,
    "launches60": 8,
    "contextMinutes": 60,
    "uniqueCreators60": 6
  },
  "extractor": "terms-v2",
  "selection": "random-eligible",
  "candidates": 326,
  "calibration": null,
  "methodology": "attention-v1",
  "termHistory": null,
  "contextCoverage": 0.9990605555555555,
  "memoriesRecalled": 0,
  "priorWindowCoverage": 0.9998434259259259
}
Expected
{
  "direction": "down",
  "magnitude": null,
  "magnitudeNote": "v0 does not model magnitude",
  "probabilityUp": 0.4829048746269394
}
Engine narrative-heuristic-v0

P(term's share of new Pump.fun launches is higher in the next H than in the previous H). Hand-set logistic heuristic over 60 minutes of launch observations: short-term momentum, creator diversity, volume. Slight mean-reversion prior because candidates are selected while already popular.

What happened, after

Graded2026-10-15 02:33 UTC
StatusSCORED
OutcomeACCELERATED
ClassificationINCORRECT
Brier0.267
Log loss0.728
Graderjudge-v1
Outcome data
{
  "shareAfter": 0.004472733528298641,
  "shareRatio": 1.2200760937703656,
  "totalAfter": 11626,
  "shareBefore": 0.003663003663003663,
  "totalBefore": 13104,
  "coverageAfter": 1,
  "launchesAfter": 52,
  "coverageBefore": 1,
  "launchesBefore": 48
}
Post-outcome analysis · judge-v1

Predicted 48% that "keep" would gain share over 6H. Its share of launches moved 0.37% → 0.45% (+22%). INCORRECT.

— momentum was right but too weak on "keep" (6H). momentum contributed +0.26 toward yes, the true outcome, but could not overcome the prior.

— The prior was wrong on "keep" (6H). Every active signal pointed the right way; the base assumption outweighed them.

2 memory records written (sealed: written while memory was short-context only).

Integrity

Hash4b7bd6696cee61ece6e00c51d973eaba12be97fbb7c9cf0100239dc1e7777d1d
Previous hash94ad957b2ef966b8597509c22c4d09b76b055997dd73873e8ba0892fdef12f09