The Validation Layer for Meme-Coin Momentum

Signal layer. Proven. Not promised.

Alpha Lumina finds early attention, rejects fake strength, explains every decision — and grades whether each signal was actually tradeable: liquidity depth, fill cost, round-trip risk. Before it ever reaches a subscriber.

Join the Waitlist See How It Works
Early Discovery Source Measured Signal Validated Tradeability Check Risk Rejected
Phantom●Raydium● Jupiter●Solana● Pump.fun●Magic Eden● JITO●Pyth● Orca●DexScreener● Birdeye●Helius● GMGN●Marinade● Tensor●Drift● Phantom●Raydium● Jupiter●Solana● Pump.fun●Magic Eden● JITO●Pyth● Orca●DexScreener● Birdeye●Helius● GMGN●Marinade● Tensor●Drift●
Signals scored
258K+
Score computations logged since launch
Signals today
—
Tokens evaluated in the last 24 hours
Score dimensions
20+
Wallet · Liquidity · Momentum · Risk · Convergence
Validation
Paper
Every signal tested before subscriber exposure
How it works

The engine is the product.

Raw on-chain data flows through a four-stage intelligence pipeline. Each step filters, scores, and validates — so only what survives reaches your feed.

01

Ingest Market Data

Pull wallet activity, token launches, liquidity shifts, and smart money behavior in real time from Solana.

→ wallet_4kQ..rM swap 2.4 SOL
→ wallet_8nP..vL add_liq 18 SOL
→ pump.fun new_token $FADE
02

Score Every Signal

AI engine ranks tokens across five dimensions — filtering noise and identifying which setups deserve attention.

wallet_quality  ████████░░ 88
failure_risk   ██░░░░░░░░ 24
→ composite: 87 → PASS
03

Route Best Opportunities

Only signals that clear scoring thresholds enter paper validation. Weak setups are filtered before any exposure.

✓ $PEPE score 87 → paper_test
✓ $AVCI score 91 → paper_test
✗ $FADE score 43 → filtered
04

Act with Confidence

Paper-validated signals delivered with full score breakdowns so you can move on intelligence, not hype.

signal: AL-2497
outcome_1h: +34.2%
status: APPROVED → feed
Technology behind the feed

What earns a place in your feed.

The feed is not random. Every signal passes through five evaluation layers before it reaches a subscriber.

alphalumina_engine · scanning
> scanning new token launches...
> analyzing wallet clusters...
> checking liquidity depth...
> scoring social velocity...
> filtering low-quality hype...
> evaluating failure risk patterns...
> cross-referencing elite wallets...
> ✓ 3 signals routed to feed
> ✓ 12 signals filtered out
> cycle complete. next scan in 10m...
Layer 01

Wallet Intelligence

Detects smart money movement, elite wallet convergence, and suspicious cluster behavior across 600+ tracked wallets.

Layer 02

Liquidity Quality

Checks whether volume and liquidity are real or fragile. Filters tokens with shallow pools and artificial volume.

Layer 03

Momentum Scoring

Measures price momentum, volume velocity, social acceleration, and on-chain activity shape in real time.

Layer 04

Risk Filtering

Removes weak, spammy, honeypot, and high-failure-risk signals before they reach the feed. Safety first.

Layer 05

AI Ranking Layer

Prioritizes opportunities by combined score calibrated against 5.9M historical outcomes. Only the strongest survive.

The method

The value is discipline, not velocity cosplay.

01 / Scored

Every setup ranked before anyone is asked to care.

Tokens evaluated across wallet quality, liquidity behavior, momentum profile, safety filters, and cross-signal convergence. Weak structure dies in the scoring layer.

02 / Tested

Paper execution before subscriber exposure.

Every candidate passes through staged paper validation. The system logs what it would have traded and tracks where ideas broke down. Misses are model feedback.

03 / Tracked

Post-signal truth is part of the product.

Win rates, runner rates, expected value — measured at 15 minutes, 1 hour, 4 hours. Published openly. Hit rate alone is lazy. Runner rate matters.

Before you ask

Questions a skeptical trader will ask.

01Is this just another alert channel?
No. The differentiator is disciplined scoring, paper validation, and visible outcome tracking. Speed without filtration is just faster bad decisions.
02Why should the score be trusted?
It is anchored to wallet behavior, liquidity conditions, and historical outcome calibration — not sentiment, narrative, or mood.
03Who is this for?
Active Solana traders who want fewer, higher-conviction setups rather than endless pings pretending quantity is value.
04What does "tradeable" actually mean here?
That a realistic position could have entered after the flag and exited into the move — net of price impact against actual pool depth, fees, and a real stop. A token that "did 10x" after round-tripping through zero liquidity was never tradeable, and we grade it that way. Peak-price screenshots are how this industry lies to you.
05What do founding members get?
Early access, locked pricing, and a front-row seat to how the system performs before broader release. Small cohort, tighter feedback, less noise.
Receipts, not promises

Real receipts from the grading engine.

Two real entries from our logs — contract addresses and dates included. One clean signal traded start-to-finish on paper. And the audit where we caught our own grader misreading a broken data feed — fixed, re-graded, and republished the same day. The countable track record starts 2026-07-04, the day the instrument passed its audit; everything measured before that is calibration, and we don't claim it in either direction. More on the receipts page.

α
AlphaLumina · Validation Log
● graded outcomes
✅ SIGNAL → PAPER TRADE — clean lane
📈 $MOO — Solana · 2026-06-29 09:53 UTC · score 64
💰 MC:$56K  Vol:$81K/h  Liq:$17K at entry
✅ TP1 hit +2.5min · TP2 +5.5min · TP3 +7.5min — laddered exits, not diamond hands
⚠️ 0.125 SOL position, sized to the pool — simulated, no real funds
🔗 gmgn.ai/sol/token/2qvi…pump
outcome: trailing stop closed the position 9m 34s after entry — +81.6% realized on paper, full lifecycle logged: entry, three partial take-profits, exit. This is what the engine does when it's right.
🔧 THE INSTRUMENT AUDIT — 2026-07-04
⚠️ found: our liquidity feed emits 0.0 when a token migrates pools — while trading continues. The grader was reading those zeros as deaths.
✅ caught by cross-checking independent price data · grader fixed · every fire re-graded · pages updated the same day · locked with a regression test
a validation layer that can't audit itself is just another feed. Results measured before this audit are calibration data — we don't count them as wins or misses. The forward track record starts here, with a proven instrument.
Full-Lifecycle Logging
Every paper trade is logged end to end — entry, each partial take-profit, the exit that closed it — so a "+81.6%" is a complete story, never a cherry-picked peak.
Tradeability Gate
"Went up" isn't enough. We model fill cost against pool depth — price impact plus fees at $500, $2K and $10K position sizes — and grade net of it.
Round-Trip Rule
If liquidity dies to near-zero after a flag, that fire is invalidated — a later pump is a new episode and never retro-credits the original call. No resurrection accounting.
Rejection Discipline
Clone traps, dirty flow, bundled launches, late attention — vetoed and published with reasons. The rejects get graded too, so you can audit whether the vetoes were right.
Founding list

Build with the first
serious cohort.

Founding spots are limited. Better feedback, tighter iteration, less noise.

No credit card · No spam · Founding pricing locked

✓ You're on the list. We'll be in touch.