EyezOn · Technical Blueprint · Discovery

The Iris Discovery Layer

Owning graduation detection — chain-wide, real-time, $0.

39azUYFWPz3VHgKCf3VChUwbpURdCHRxjWVowf5jUJjg

One account. Pump.fun's migration authority signs every graduation on the chain — ~80–150 per day, all flowing through this single point.

The Verdict

Build our own chain-wide graduation detector on free public RPC.

$0/mo, no paid Helius, reuses code we already run. Watch one account — pump.fun's migration authority 39azUYFWPz3VHgKCf3VChUwbpURdCHRxjWVowf5jUJjg — which signs every graduation chain-wide (~80–150/day). Poll it every 10–15 seconds.

$0/moRunning cost
6–20×Faster than today
10–15sPoll cadence
ZeroNew infra
The Problem

Today the scanner discovers tokens via GeckoTerminal's new_pools feed — a shared, third-party pipe that everyone (including Quantix) drinks from. It runs ~60–180s behind the chain and is rate-limited. We don't own our own eyes.

Architecture Decision

Four ways to watch the chain

All four were weighed on latency, cost, and failure behaviour. One wins now; one wins later; together they're the plan.

Option Latency Cost Reliability Verdict
aHelius Enhanced Webhook address-scoped push ~1–3spush Free tier~4.5k credits/mo Push-only — silent gaps if used alone Upgrade layer, later
bGeyser gRPC LaserStream <1sstream $499/momainnet Heavy stream for one account Rejected — overkill
dHybrid c now · a later · c always the safety net Best of both $0until webhook opt-in Highest — push speed with poll-backed coverage The real recommendation
Key Addresses

Three coordinates on the chain

Migration authority — the signal 39azUYFWPz3VHgKCf3VChUwbpURdCHRxjWVowf5jUJjg
Pump.fun bonding curve 6EF8rrecthR5Dkzon8Nwu78hRvfCKubJ14M5uBEwF6P
PumpSwap AMM pAMMBay6oceH9fJKBRHGP5D4bD4sWpmSwMn52FMfXEA
Phased Rollout

Five moves, each one safe

Standalone first, shadow-proven second, production last. Nothing touches the live scanner until the numbers say so.

1

Free-RPC graduation poller — standalone

Build the poller as its own process watching the migration authority. No production surface touched.

2–3 days $0 No sign-off needed
2

Shadow mode — alongside Gecko

Run both feeds in parallel and prove 100% coverage plus the latency gain with 3–5 days of real data.

1 day $0 3–5 days of proof data
3

Cutover into the scanner

Swap the scanner's discovery source to the poller once shadow numbers hold.

0.5–1 day ⚠ Needs Pete's sign-off — touches production
4

Full Iris handoff — zero Gecko dependency

Retire the third-party feed entirely and re-verify the backtest gate against the fresh-graduation profile.

2–3 days $0 Backtest gate re-verified
5

Helius webhook upgrade — optional, → ~1–3s

Layer address-scoped push on top only if speed becomes a competitive edge. Poller stays on as the safety net.

Only if speed matters Free tier
Why It's a Moat

Rivals all read the same shared feed

Real-time push, not poll

Events arrive as they happen once the webhook layer lands.

Zero rate limits

Our cadence, our call volume — nobody throttles us.

Zero third-party dependency

No shared pipe that can slow, cap, or cut us off.

Un-copyable

Rivals all read the same shared feed; this edge lives in our own plumbing.

Composes with our scoring moat

Authenticity + whale + crowd signals fire on fresher targets.

Bottom Line

~1.5–2 weeks phased. $0/mo throughout.

Every phase is reversible, every dollar stays in our pocket, and the eye we build is ours alone.

Next Step Build Phase 1 — standalone, no risk, no approval needed.