METHOD · pulse-risk/2
How Pulse decides
The score is computed by fixed rules, not by AI. The AI only explains the evidence, and a verifier checks every claim it makes before you see it.
01Read the chain
Mint and freeze authority and Token-2022 extensions come straight from Solana RPC. When Jupiter disagrees with the chain, Pulse says so and uses the chain.
02Measure the exit
Pulse asks Jupiter for real sell quotes of $100, $1k and $10k into USDC. Quotes only: Pulse never sends a swap.
03Check the creator
How many tokens the creator wallet has launched, how many graduated, and how much it still holds.
04Explain, then verify
The AI writes a brief from the evidence ledger only. Claims without evidence, invented numbers and advice wording are removed.
THE TEN FACTORS · WEIGHTS SUM TO 100
Mint authority18Solana RPCActive = 1, disabled = 0. Active is a critical flag.
Freeze authority10Solana RPCActive = 0.8. Active is a critical flag.
Token-2022 extensions10Solana RPCWorst extension wins: permanent delegate, non-transferable or frozen-by-default = 1 (critical); transfer hook or pausable = 0.7; transfer fee scaled by basis points.
Holder concentration16RPC, owners resolvedLargest wallet and top-10 wallets, with pools and curves excluded (heuristic). Falls back to Jupiter's figure, capped at 0.5.
Exit liquidity18Jupiter quotesPrice impact selling $1k: >10% = 0.9, 3–10% = 0.5, 1–3% = 0.2. No sell route = 1 (critical). Bonding-curve tokens are not scored.
Liquidity vs valuation6JupiterLiquidity under 0.5% of FDV = 0.8; under 2% = 0.4.
Age4JupiterUnder 24 hours = 0.7, fading over 20 days.
Identity4JupiterFlagged suspicious = 1, unverified = 0.4, verified = 0.
Creator launch history8Jupiter auditTokens launched by the creator: 50+ = 1 (critical, serial launcher), 10+ = 0.7, 3+ = 0.4.
Creator holding6Jupiter auditCreator's share of supply: >20% = 1, >10% = 0.6, >5% = 0.3.
Score = 100 × Σ(weight × severity) ÷ Σ(weight of factors with data). Missing data never lowers the score; it lowers coverage and confidence. Any critical flag lifts the score to at least 50. Coverage under 60%, or under 80% with a calm score, shows “Insufficient data” instead of a reassuring band.
AI CLAIM VERIFIER
- Every finding must cite an evidence ID that exists.
- Every number must match a value in the cited evidence.
- No buy, sell, “safe”, “guaranteed” or price predictions.
- No links or markup. Token names are treated as untrusted data.
- One automatic repair attempt, then a labelled rule-based fallback.
LIMITS, STATED PLAINLY
- Holder owner classification is a heuristic.
- Jupiter data is third-party; quotes are point-in-time.
- Bonding-curve tokens can't be routed by Jupiter yet, so their exit depth isn't scored.
- Pulse runs while it's open; there's no background monitoring.
- A low score is fewer warning signs, not protection. Not financial advice.
Ready? Open the launch radar →