Liquid Hacks

Liquid Hacks is a grey market for API credits. Won API credits you'll never burn? Flip them. Need cheap compute for your side project? Buy from other hackers. Direct deals, no paper trail

Comment

Inspiration

Hackathons distribute millions of dollars in API credits annually. The majority expire unused — not because winners lack ideas, but because the credits are locked to specific providers that don't match their actual stack. Meanwhile, other builders pay full price for the same compute. This is a pure allocation inefficiency: the supply exists, the demand exists, and no exchange mechanism did.

What it does

LiquidHacks is a grey market for hackathon API credits. Users can:

  • List credits for sale or post buying requests across providers including OpenAI, Anthropic, Google Cloud, AWS, and Vercel
  • Browse and filter by provider, credit type, price, and availability
  • Contact sellers directly via built-in chat without revealing contact info until a deal is agreed
  • Verify trade history — completed trades are recorded as graph relationships in Neo4j, giving users a verifiable reputation score
  • Get AI-assisted listing descriptions via Yutori's n1 model, which generates accurate, factual descriptions based on listing metadata

How we built it

Backend: TypeScript + Hono (lightweight Node.js framework), Drizzle ORM, PostgreSQL via Render managed Postgres, JWT auth via httpOnly cookies, GitHub OAuth. Frontend: Vanilla HTML/CSS/JS. No framework overhead — fast load, easy to audit.

Neo4j Aura: Each completed trade writes a (User)-[:TRADED_WITH {provider, amount, timestamp}]->(User) edge. Profile pages query the graph for trade count, surfaced as a "verified trades" trust score. Yutori n1: Browser agent called at listing creation. We pass the form metadata as a page context, and n1 returns a type_text action with a suggested description. We parse the tool call and populate the textarea. Growth tooling: Python scraper collected 518 hackathon projects and 584 contacts with GitHub, LinkedIn, Twitter, and email. Admin panel surfaces contacts with outreach tracking.

Challenges we ran into

Fraud prevention is hard! Had to think a lot about what our responsibility is in case of fraud.

Accomplishments that we're proud of

  • Functional end-to-end marketplace with auth, listings, chat, and trade tracking
  • Live Neo4j trust graph with real data model
  • 1125 contact growth pipeline with admin outreach tracking built in
  • Clean economic model: sellers recover 40–60% of expiring asset value; buyers access compute below market rate

What we learned

Secondary markets for digital assets are underbuilt. The tooling for physical goods resale (eBay, StockX) is mature; the equivalent for API credits, software licenses, and promo codes barely exists. The demand signal from even a small hackathon deployment validates the model. Neo4j's graph model is specifically well-suited to reputation systems — querying second-degree trust relationships or detecting anomalous trade patterns is natural in Cypher and awkward in SQL.

What's next for Liquid Hacks

  • Automated outreach to the 584 scraped hackathon winners to seed supply
  • Credit verification — balance proof before listing goes live
  • Smart contract escrow for high-value trades
  • Expanded Neo4j trust graph — second-degree reputation scoring, fraud pattern detection
  • Broader asset types — referral credits, sign-up bonuses, software license transfers

Yutori's n1 model was a fun puzzle — it's a browser agent, not a text model, so it returns type_text tool calls instead of completions. We had to parse the browsing actions to extract the description text it "typed." Actually kind of beautiful once you understand the model's intent.

Accomplishments that we're proud of

The economics work. Sellers offload value they'd lose anyway. Buyers access compute at 40-60% below face. The market clears. That's genuinely good for the hackathon ecosystem — it routes resources to the people most likely to use them.

What we learned

Grey markets reveal what people actually value. The fact that hackathon credits are tradeable at all is a signal: the credits have real utility, just mismatched with who holds them. Secondary markets are often the truest price discovery mechanism.

What's next for Liquid Hacks

Automated outreach to hackathon winners (we have 1125 scraped contacts ready). Escrow via smart contract for high-value trades. Credit verification, proofs of balance before listing. Expanding to non-hackathon credits (referral bonuses, sign-up credits, promo codes). And deeper Neo4j: second-degree trust scores, fraud pattern detection, reputation staking.

Built With

Share this project:

Updates