The brain never clocks out.

Every token minted gets watched. Every reading gets scored against what happened next. Every resolution re-weights what matters. This page shows the current state of that machine - not a pitch, the actual learned state as of 2026-09-19 01:20 UTC.

1,253,618coins watched all-time

425,738curve readings taken

51,357creator wallets indexed

49,722outcomes fed back into weights

What the data taught us this month

Base: 370 of 32,381 sampled tokens (1.14%) kept attracting buying between minute 5 and minute 30. Every early signal is judged against that base. Lift above 1.0x means the signal predicts continued buying; below 1.0x means it predicts the graveyard.

Signal at 5 minutes	Samples	Kept buying	Rate	Lift vs base
over 25pct bought	2,091	152	7.27%	6.4x ▲
read as moved unverified	4,897	257	5.25%	4.6x ▲
10 25pct bought	1,201	59	4.91%	4.3x ▲
1 10pct bought	6,559	98	1.49%	1.3x ▬
nobody holding	32,381	370	1.14%	1.0x ▬
read as trickle	13,862	70	0.50%	0.4x ▼
read as nobody bought	13,622	42	0.31%	0.3x ▼
under 1pct bought	8,913	17	0.19%	0.2x ▼

The headline finding: when a quarter or more of the floating supply has already been picked up by separate wallets inside five minutes, those tokens go on attracting more buying at 6.4 the base rate - across 2,091 samples. Meanwhile trickle entries and nobody-is-buying starts land at or below half the base rate. Attention follows commitment, not chatter. That gradient is what the desk trades on.

Live: the quiet-swarm watchlist

Our own finding says hype anti-predicts graduation - so the desk watches the opposite: fresh LOW-notability swarms, the cohort that graduated at 2.87 percent versus 1.08 for loud ones. The most recent quiet commits still awaiting their verdict:

Token	Notability	Committed
No quiet swarms pending right now.

Current feature weights (what the algorithm trusts today)

Feature	Weight
creatorBestAth	48
curveVelocity	33
socialTag	30
narrativeTag	8
notability	1

Weights shift automatically as outcomes resolve - creator history and curve velocity currently dominate, exactly as you would expect from the lift table above. When the market changes, the weights drift with it. No committee decides this; the resolutions do.

Why you can trust any of this

Signals under 100 samples are not reported at all - small samples dress noise up as discovery, and this industry has enough of that. Everything is measured in-sample against this feeds own window, stated rather than hidden. And none of it is a price prediction: kept buying is a fact about wallets, not about anyones profit.

Point the brain at your wallet → audit the prediction record

Not financial advice; memecoins can go to zero regardless of any signal. Assume you can lose everything you put in. Page regenerates from the ledgers - last build 2026-09-26 15:41 UTC.