Skip to content
vibestarter$ vibeathon --edition 01

Winners

[ the vibecoding championship · edition 01 · judged ]

Two weeks to take an idea to a shipped, working app, built end to end with AI coding agents. Every score below is the aggregate of the full judging panel, and every builder who shipped takes home Starter Card points.

// edition_01 poster: THE FIRST CURSOR

ED.01

// the_panel

Serpinxbtwillyham.ethCyphrClaude (AI)AIGPT (AI)AI

// the_rubric

Execution 30%Launch Potential 25%Originality 20%Craft & UX 15%Vibecoding Leverage 10%
Corvo Edge screenshot1ST PLACE$500 · FUNDED LAUNCH · 300 PTS

Corvo Edge

by Smartcoded (@smartcoded)
75.1panel score / 100

A chat-first terminal that folds swaps, tokenised equities and FX across four chains into a single self-custody interface. The panel's clear winner — the most complete product in the field, and the most ambitious.

built with · Claude Code (Claude Opus 4.8)

An unusually ambitious consolidation of multichain trading into a coherent, polished chat-first interface.
GPT (AI)
ONETOKEN screenshot2ND PLACE$250 · FUNDED LAUNCH · 200 PTS

ONETOKEN

by OneToken
73.5panel score / 100

A permissionless launchpad on a one-sided Uniswap V4 curve — launch with no seed liquidity, trade everywhere. The best-crafted build of the field.

built with · Dev

Looks well built and nicely designed. Worked quite well — launching on a one-sided curve with no seed liquidity is a clear wedge.
willyham.eth
aHand screenshot3RD PLACE$100 · 200 PTS
72.9panel score / 100

An onchain trust network that rewards the whole chain of people who found your answer, not just the solver. Judges rated it the most original idea in the competition.

built with · Claude (sponsored Vibeathon Claude Pro seat); builder is a full-stack dev since 2007 and multi-time ETHGlobal finalist

An original positive-sum referral protocol whose incentives reward both the solver and the human chain that found them.
GPT (AI)
04FutureRootsby CSH
66.3panel score / 100

Family vaults for memories and generational wealth, built with unusual compliance rigour.

built with · Claude Code, Fable 5 & Opus 4.8

Production-grade scope shipped for real. The hardening and compliance focus is unusual and commendable for a hackathon.
Claude (AI)
05exSafeby exsild
65.3panel score / 100

Paste any link, contract or transaction; get one plain-language verdict before you sign.

built with · Claude Code

Turns complex wallet-security signals into one clear, actionable community-facing result across multiple surfaces.
GPT (AI)
06EventChainby Muktar
60.4panel score / 100

End-to-end event management with QR attendance verification, working today.

built with · Next.js, React, TypeScript, Prisma, Supabase, NextAuth.js, PDF-lib, QRCode, Solidity, Hardhat, Base Network, Vercel, and AI agents (Claude Sonnet 5).

Works decently well. Not particularly novel, but a genuine product — a complete organizer-to-attendee workflow. Solid start.
willyham.eth
07Lineageby @EdCriptoFi
56.2panel score / 100

Shareable onchain provenance cards pairing Ethos reputation with ERC-8004 agent identity.

built with · Claude Code

Makes human and AI-agent provenance legible as a compact certificate, aligned with the trust tiers Vibestarter already talks about.
GPT (AI)
100 STARTER PTS
08Grayl AIby Ibrahweb3
55.5panel score / 100

AI risk scoring for NFT mints across nine chains, with reputation-gated access.

built with · Claude(Fable 5), Figma, chatgpt(GPT-5.5)

A polished, information-dense presentation of NFT scoring, risk signals, whale tracking and reputation-gated access.
GPT (AI)
09Agent Verseby Kelly / @kellyoliseh
51.0panel score / 100

A marketplace for AI agents with sandboxed runtime and scoped permissions.

built with · Claude code

A polished and unusually complete marketplace shell, with clear product taxonomy, permission disclosures and publisher economics.
GPT (AI)
10AIDOLby Towertech_
47.9panel score / 100

999 AI agent NFTs on Base, where traits shape how each one talks and behaves.

built with · CLAUDE CODE

Makes NFT traits govern an agent's personality and behaviour rather than serving only as visual metadata.
GPT (AI)
100 STARTER PTS

[ vibeathon_02 ]

This happens regularly.

$ next edition: entries opening soon — ship your build in the next one.

>_ Enter the next Vibeathon

// Every score is the aggregate of the full judging panel: the mean weighted score across all judges who scored the entry, snapshotted at finalization. Rubric weights above.