The strict gate. Every candidate is verified before it’s allowed on the list — anything that fails a hard check is dropped, not flagged: active mint authority, honeypot or sell tax over 10%, LP verified unlocked, dev wallet holding over 3%, or still on a launchpad bonding curve. Survivors must also clear your liquidity, volume, holders, mcap, and minimum-score bars. Website / X / Telegram presence earns a small bonus and shows as links. “Unverified” tokens (no checker covers them) aren’t auto-dropped but score 0 on safety, so they rarely clear the bar — treat any that do with extra suspicion. Volume gets sanity-checked too: 500+ transactions averaging under ~$12 each is wash-bot traffic (thousands of dusted “buyers” faking demand), flagged on every row and dropped outright when blatant. Contract checks can’t see sell pressure: a minutes-old token can pass every safety gate and still get nuked by launch snipers dumping in unison — that risk lives in age, so the age gate (default ≥6h) keeps you out of the window where most coordinated dumps happen, and anything under 6h that does show carries a “launch dump window” flag. Every survivor also gets an ATH check from its pool’s full daily history: “−X% from ATH” shows how far it’s pulled back, 70%+ is marked as revival territory, and the retrace filter can show only deep pullbacks — tokens that already ran, crashed, but still pass every safety gate are a different bet than fresh launches: the rug risk is behind them, the question is whether the community is. (ATH is measured from the current main pool — a token that migrated pools early may read slightly low.) Passing the gate means “worth your time to research,” never “safe to buy.”

Rules, executed exactly: hard stop at your −%, sell half at your take-profit multiple, and once a trade is up 50%+ a trailing stop rides 20–35% below its peak — plus an instant full exit if the pool loses 70% of its liquidity (rug). Starts with a fake $1,000 bank, max 5 open trades. Prices update every minute while the app is open; nothing trades while it’s closed, and paper fills ignore slippage and fees, which on thin memecoin liquidity are real — so treat these results as the strategy’s best case. If the bot can’t beat the bank on paper over 20+ trades, the live version would only lose money faster. This is the evidence layer, not a money printer.

Pulls each watched Solana wallet’s recent parsed transactions through your Helius key and keeps only real buys — token in and SOL/tokens out in the same transaction, so airdrop spam doesn’t flood the feed. Anything newer than your last check gets a NEW badge; with alerts enabled and the app open on a computer, new buys fire a browser notification (phones only show badges — real phone push needs a server, ask me when you want that). Auto-check polls every 5 minutes and spends Helius credits per wallet each pass, so leave it off when you’re not actively hunting. Tap Gate on any buy to run that token through the Scanner’s full gate before you even think about following the trade. Every buy it sees is also logged as that wallet’s “call” with the mcap at the time, then re-checked on later runs — after a few days each watched wallet shows its own scorecard above: how many calls hit ×2+, average peak, best call, typical entry size, and how many went to zero. Calls need 6 hours before they count, so scores fill in gradually. That’s how you find out whether a wallet is actually worth following instead of guessing.

How discovery works: it pulls a winner’s top holders straight from the chain, drops pool and locker accounts, and marks bot-frequency wallets and freshly created wallets. Every run is remembered — a wallet that shows up across two or more of your winners gets a Repeat badge, and that overlap is the real “finds gems early consistently” signal. True win-rate and average-multiple stats need full trade history that free APIs don’t expose, so vet shortlisted wallets on their GMGN page (free — shows PnL, win rate, avg hold time) before you Watch them. Discovery finds candidates; GMGN proves them.

The first minutes of a launch, ranked by volume velocity (dollars per minute since the pool opened). Volume alone lies — so every row shows unique buyers vs. total buys and average trade size from the last 5 minutes: 900 buys from 40 wallets at $9 each is one bot; 300 buys from 260 wallets at $80 each is a crowd. BOT VOLUME and ONE-SIDED (buys with almost no sells — someone controlling the float) flag the fakes. “Insiders?” runs the full RugCheck pass on demand for bundle/top-10 data. Trade links open Jupiter or pump.fun with the token loaded — you sign in your wallet. Nothing here passes the age gate; this is the scalp table, not the investment table: tiny size, exit into strength, and remember the tool sees a spike 30–60 seconds after the bots did — except the LIVE section, which streams straight from the chain via PumpPortal (free, third-party): new tokens appear the second they exist, buyers and velocity update per trade, and every watched wallet’s pump.fun buys and sells show as they happen, with a WALLET IN tag on the launch row and a notification when a graded wallet steps in. Those live buys also feed the scorecards with exact entry mcap. Stream runs while this tab is open with “live” on.

What one cycle does, in order: refreshes the Scanner and auto-tracks anything scoring 80+ so its outcome gets recorded → re-checks every tracked token’s outcome → for each new WINNER (peaked ×2+), pulls its top holders on-chain and walks back to its first 30 buyers, checking who has already exited (≈300 credits) — early-and-out is the sharp-money profile — and logs them as candidates → any candidate seen early in two or more different winners gets auto-watched → polls watched wallets for buys and sells, updating scorecards, hold styles, real SOL profit and who-moves-together → checks the SOL balance of every top holder on up to 5 new gate-passers (≈3 credits each) so a “Whales: 3 hold $50K+” chip shows — and every real top holder of those gate-passers (anyone under 15% of supply) is logged as a candidate, so any wallet showing up in two different gate-passers gets auto-watched without waiting for a token to double → re-counts holders on every tracked token (free) so ACCUMULATING / PROPPED / BLEEDING trends show on the Watchlist → traces who first funded up to 10 untraced wallets (≈20 credits each) so funding rings surface → retires auto-added wallets that grade D after 5+ calls or go silent for two weeks, keeping them on a retired list so the track record survives. Runs only while this tab is open in a browser — leave a laptop on with the app up and it works the whole time. Credit cost, honestly: every wallet poll and every holder profile is an Enhanced call at 100 credits, so 25 wallets per cycle = 2,500 credits, and mining one winner ≈ 1,500. On the free tier’s 1M credits, hourly cycles with 25 wallets last about two weeks; 30-minute cycles about a week. The Developer plan ($49, 10M credits) runs 30-minute cycles for two-plus months. Wallets you added yourself are never auto-retired.

Automated meta detection: every scan pulls hundreds of trending + freshly launched pools (Solana, BNB, Robinhood), splits their names into words, and counts how many distinct tokens are riding each word right now, with combined 24h volume. When devs pile onto a theme, the word count spikes here within minutes — this reads what’s being launched and traded, which usually front-runs what’s being tweeted. ↑ and NEW badges compare against your previous scan. Tap a word to see every token using it.