Photo of my cofounder and I back when we were just two college kids at Berkeley.
I was on track to graduate with a double major in CS and Math, and had multiple full-time offers lined up.
Then we got into @alliance.
We decided to go all in on building @predexon.
I dropped the math major and graduated that semester with just a CS degree.
School teaches you to maximize optionality.
Startups are about deliberately giving most of it up.
Told my agent to place a $1 bet on Polymarket.
It bought data from @teneo_protocol to make a decision, then placed the order via the @predexon agent on Teneo. All without any UI or human input.
a lot has landed on Predexon lately.
⏱️ pending trades reach our stream 3–5 seconds before first on-chain confirmation. The fastest way to track Polymarket flow.
🔎 market discovery + orderbook history now span the major prediction market venues: @Polymarket, @Kalshi, @trylimitless, @opinionlabsxyz, @predictdotfun
🌐 50,000+ markets through a single websocket. full depth orderbooks, snapshots on subscribe, live deltas, trades, tick size updates. same latency as Polymarket’s native orderbook websocket.
🚨 new channel: oracle events. proposals, disputes, settlements, resets. follow market resolution in real time.
🧩 trading API is live for Polymarket and Predict.
📊 some of the deepest smart money analytics APIs in prediction markets, powered by billions of rows of stored and computed market, trade, and wallet data.
🚀 more coming soon: low-latency trading across the major venues, plus bigger infra pushes behind the scenes.
one layer to stream, query, and trade across prediction markets.
Hot take:
In the next two years you’re going to see a Prediction Market TOOLING startup reach unicorn status.
Give users what they want.
That’s worth billions.
Spent part of the weekend trying to make a pitch deck with AI.
Searched “AI presentation,” tried a bunch of the tools that came up: Genspark, Manus, and others.
Honestly, most of them felt the same. The output looked obviously AI-generated, overly polished in the wrong way, and kind of vibe-coded. Hard to tell what the real product differentiation even is.
In the end, I just went back to Figma and made the deck myself.
AI is getting good at a lot of things, but for something as high-stakes and taste-driven as a pitch deck, it still feels very far from replacing real judgment.