Earlybird

Inspiration 💡

You know the feeling. You saw the trend before anyone else. You called the sound, the creator, the format before your group chat, before your FYP caught up, before the internet noticed.

And you got nothing for it.

Platforms reward you for making viral content, but there's no market for predicting it. Your cultural instincts are a real skill, and that skill has no price.

Earlybird is a prediction market that pays you for being right before everyone else.

Why this, why now 🔥

Three things converged that made Earlybird possible this year.

Neural models can predict virality before humans see content

Meta's TRIBE v2 is trained on fMRI data. It predicts how a human brain will respond to a video, which means we can generate opening odds the instant a reel is uploaded, before it gets a single view. No prediction market has ever had a cold-start oracle like this.

Solana makes micro-stakes betting economical

A bet costs fractions of a cent and confirms in under a second. Scrolling and betting at TikTok speed only works on Solana-tier throughput.

Cultural taste is becoming a measurable asset class

People already invest attention in reels, hours of it, daily. The question "is this going viral?" is the most-asked question of our era, and it's never had a market.

What it does 🎯

Earlybird is Polymarket for internet virality. You scroll a TikTok-style feed, and every reel is a live prediction market on whether it'll hit viral thresholds in 48 hours. One tap places a bet. The earlier you call it, the more you win.

TRIBE v2 scores the reel on attention, arousal, novelty, and retention. Our formula turns those into opening odds, with retention as a multiplier (a reel no one finishes can't go viral).
Users stake $STAKE SPL tokens on Solana devnet. Every bet is a real on-chain transaction with a Solscan receipt.
MongoDB change streams propagate odds in realtime. Bet on the phone, watch the web leaderboard react within 300ms.
Vector Search clusters users into tribes based on betting history. Tribes that disagree on a reel get surfaced aggressively. Disagreement is the engagement engine.
Winners get earliness-weighted payouts. Calling a reel at 35% before it hits 1M views pays more than calling it at 80% after the trend is obvious.

How we built it 🛠️

Mobile (React Native + Expo): scroll feed, burner wallet, betting. Solana keypair auto-generates on first launch and persists to device keychain. No wallet install, no seed phrase, no friction.

Web (Next.js): landing page, live leaderboard, and the neural activity monitor, a real visualization of TRIBE's four score dimensions as a brain overlay.

Solana (devnet): custom $STAKE SPL token, user-signed transfers on every bet, admin-signed payouts at resolution.

MongoDB Atlas: documents for reels/bets/users, change streams for realtime odds, and Atlas Vector Search for tribe clustering (users as embeddings, tribes as centroids, cosine similarity reassigns in 200ms).

TRIBE v2: off-chain neural oracle. Predictions post on-chain as the opening line, and the market clears on-chain.

Challenges we ran into 😵

TRIBE v2 was the hardest piece. Running Meta's model on short-form vertical video required dependency surgery; it wasn't designed for 30-second reels.

Expo's dev build for Solana is brittle. Native crypto polyfills have to load in a specific order or Keypair.generate() silently produces weak keys.

Connecting mobile to our local backend through hackathon WiFi was a saga. We ended up running everything through personal hotspots.

Accomplishments we're proud of 🏆

We shipped a real product, not a slideshow. Every bet is a real Solana transaction you can verify on Solscan. Mobile and web sync live through MongoDB change streams: bet on the phone, watch the leaderboard react within a second.

We're proud of the product identity. Prediction markets usually look like spreadsheets. Earlybird looks like something you'd actually scroll, with serif italic market copy, terminal-blue data, and yellow only appearing when real money is on the table.

The idea earned its name. We iterated through a lot of framings (content betting, creator tokens, attention futures) before landing on earliness as the thing being rewarded. That reframing is what made it click.

What we learned 📚

Prediction markets are a language for belief, and applying them to cultural foresight (something people already do in group chats and FYPs) unlocks a use case sitting in plain sight.

"AI plus blockchain" is usually a buzzword stack. But when AI produces off-chain signal (TRIBE scoring a reel) and the blockchain settles on-chain consequences (payouts to winners), the combination stops being a gimmick and becomes an architecture.

Hackathon demos live or die on moments where the tech stops being invisible: the Solscan link opening, the balance jumping up, the leaderboard updating live. Make the receipts loud.

What's next for Earlybird 🐦

We want Earlybird to become the market layer for internet attention itself.

Near-term: expand beyond seeded reels to live YouTube Shorts, TikToks, and Instagram Reels as they post. Ship a mainnet version through a licensed offshore entity, keeping devnet as a no-stakes training ground.

Longer-term: creator-side markets (bet on a creator's next reel), collaborative tribes (bet and win as a collective), and a reputation API so other apps can query "is this user in the top 5% of early-viral callers?"