We watch the smart money so you don't get rugged
When proven wallets ape a fresh launch early, you see it. When it's a cabal bundle dressed up as organic demand, we flag it. No black box, no hopium, just the signal we can actually measure.
Broader market context sits in Atlas: public briefs for SOL, BTC, VIX, and the full Atlas symbol directory.
Narrative Confluence · read-only
Narratives, tokens, and wallets colliding
The narrative engine joins fresh token metadata, live Pumphouse wallet evidence, funder context, and a public-source narrative vocabulary. Public rows are delayed and redacted; the full wallet and mint trail is kept behind gated access.
Narrative radarvelocity, acceleration, authenticity, and matched fresh tokens
Confluence boarddelayed public hits where narrative and wallet activity overlap
Wallet Findercopy candidates ranked by narrative overlap and lane evidence
Brain status / overlay evidenceshape-paper overlay, drift replay, substrate, and live safety boundary
Real-money execution · runtime-gated
Live cockpit
Live execution is controlled by the runtime gate, not the static page copy. This panel shows the real wallet, held bags, armed state, stake, cap, drift, and exact executor state from the live feed.
Execution feedrecent entry gates and fills
Trending intelligence · read-only
Tokens heating up, wallets arriving early
This board does not fire trades. It tracks live token velocity and shows the wallets that bought before or near trend ignition, so we can promote only wallets that repeatedly catch runners before the crowd.
Trending tokensscore, mcap move, tape, and early buyers
Wallets to profilepresent before trend, not promoted automatically
🎯 Runner lanes · paper research
The new lanes, hunted from real runners
Every lane roster comes from the runner hunt: wallets that were repeatedly in the first 20 buyers of tokens that went on to $100k or more, verified against real post-migration market caps. CORE follows the persistent multi-runner catchers. MONSTER follows wallets seen inside the launch crews of several $1M+ tokens. FRESH follows brand-new wallets with flawless early records, expected to churn fast. C re-derives itself daily from live hit rates. Standard paper stake £0.50 per snipe; momentum lanes use £2.50, no real money.
Lane positionsevery paper snipe in the selected lane
Transaction historysnipes, trims, and closes in order
Live signal feed
Proven wallets, buying early
Each row is a wallet from our tracked alpha cohort that bought a fresh token early in its life. Higher follow score means a stronger track record. Refreshes automatically.
Alpha wallet rowstracked wallet buys, hidden until opened
Fresh launches
Every new launch, verdict-stamped
We grade recent pump.fun launches as they appear. Green means at least one of our proven wallets bought it early, that is a launch we snipe, bundled or not, because the proven money is in. Red means a bundle or spray with no proven wallet behind it, so we pass. Amber means no proven wallet in the early buyers yet, we keep watching. Every verdict is provisional and updates as more buyers land.
Launch cardsverdict-stamped, newest first
The edge, honestly
What we measured, and what we won't pretend
Black-box scanners sell certainty they don't have. We'd rather show you the homework. This is the moat: an honest edge you can pressure-test.
The homeworkmeasured claims, admitted limits, and what we just fixed
✅ What we measured
- ≈3.5×more likely to survive when a proven wallet buys early, versus a random fresh launch.
- 0.75ranker AUC on cabal-avoid: bundle launches are separable from organic ones.
- 329tokens ran past $100k in the last nine days, five past $1M. The velocity lanes copy the wallets that kept showing up early in those.
- 31×in 70 seconds, on a real tracked trade: one wallet bought a fresh launch and sold the whole bag into the first spike, then it bled to dust. The lesson we built on is that the edge is the exit, not just the entry. The VELO lane sells the instant a coin doubles.
⚠️ What we won't pretend
- 8dSurvival numbers come from one 8-day training window so far, not years of data.
- ≠Early on paper is not copied profit. Latency and fees eat edges; lanes earn promotion only after beating both.
- 2×VELO sells the whole bag at 2×, so by design it cannot catch a moonshot. It trades the rare 100× for a high hit-rate. The new velocity-runner lanes are our attempt to fix that, and they are only days old.
- ?Some big winner marks cannot be replayed: the listener kept price ticks selectively. We now persist ticks for every open position, so every future result is fully checkable.
🔧 What we just fixed
- +4new velocity-runner lanes built: bank most of the win at 2×, then ride a slice for the rare 30×+, with a hard time cap so a bag cannot rot to dust. Running A/B against the originals now.
- −3persistently losing lanes retired. The board only shows what is actually working.
- £banked winners were rendering below their entry price, a display bug where the money was always real. Fixed, so a profit can never look impossible on the board.
- 1×one lane was an overlay of another and was being double-counted in headlines. Now labelled "overlay, not additive", so nothing is inflated.
The cohort
The wallets we follow
Every wallet whose early buys we watch, including the four lane rosters above and the research watch tier. Re-derived on live data, so it changes over time.
🔑 Full wallet list is token-gated