The problem Poké solves

The Problem

Web3 today forces you to trust things you can't verify. Checking a wallet, exploring a transaction, and running a suspicious link through an AI "scam checker" all live in separate tools — and when any of them use AI, you get back a bare risk percentage with zero evidence behind it. You're left with two bad options: blindly believe a black-box verdict, or dig through raw block data yourself with no guidance at all.

The Whistleblower Gap

It gets worse if you're the one who actually catches fraud on-chain. There's no safe way to prove what you saw and when you saw it without exposing your identity — so people who spot real fraud are stuck choosing between staying silent or putting a target on their own back. That's a gap nobody in this space has meaningfully closed.

Why Poké Is Different

Poké closes both gaps at once, and does it by changing what "trust" means rather than just adding another AI feature on top. Scam Detection doesn't hand you a bare score — it shows the evidence, the reasoning trace, and a Gonka Request ID you can independently verify against Gonka's own records, so you're checking the verdict instead of believing it. Whistleblow doesn't ask for a name — it anchors a cryptographic commitment of the report and evidence on-chain, so the timing and content can never be disputed later, and the person behind it never has to be identified.

The Net Result

One command center where inspecting a wallet, judging a scam, and proving fraud all work the same way: nothing asks for your trust, everything hands you a receipt.

Challenges we ran into

Our main hurdle was GonkaRouter's 30-second inference timeout. Scam Detection and Whistleblow both ask Gonka to reason over real evidence, and with dual-model consensus running two inferences at once, that timeout shaped how we built every request in both features.