making machines dance | (F/ai - F26)
Notes on building an intelligence that gives attention back, starting on a fuel forecourt. We have been building Manako for months. This week I wrote down the thing we should have written first. Not
computer vision is not solved. not for real-world deployment, not at scale, not horizontally across the environments that actually matter. what we have had for twenty years is a series of narrow wins
nobody fully understands why the golden ratio works. it shows up in nautilus shells and renaissance paintings, in the parthenon and the iphone. designers reach for it intuitively. when a composition
Most founders are reacting to agents like it’s a UX shift. It’s not. This is an economic shift. The “product” is no longer an interface you persuade a human to adopt. The product is an output that can
TL;DR: The benchmark paradox is simply this: benchmarks guide progress, but real deployments reveal meaning. And in Vision AI, where the world is messy, unstructured and endlessly surprising, the only
For years, decentralised AI lived in R&D circles, powerful in theory, promising in benchmarks, but rarely deployed inside traditional enterprises. The gap wasn’t the technology. It was the delivery
We didn’t start by trying to fix computer vision. We started by trying to fix sports. The goal was straightforward: build the intelligence layer that clubs, players, and analysts never had access to.
My humble research contribution to @webuildscore has long revolved around vision intelligence, teaching machines not just to see, but to understand what they’re looking at. The idea that vision can
How to Tell Which KOLs Are Providing Alpha and Who's Selling a Story I've been in crypto long enough to lose count of the cycles and slogans. The one that stuck, "this is an attention market", is both
The signal isn’t visual, it’s temporal. Frames are postcards. Markets need sentences. When Kalshi popped up in a South Park episode, it was a wink that prediction markets have gone mainstream. Now
the premise The next wave of market data is not a spreadsheet; it is a video feed. Treat cameras as sensors, not scenery, and the world resolves into a high-frequency time series you can measure,
In prediction markets, speed and calibration win. Images contribute to both, by revealing what numbers miss. In “MM DREX: Multimodal Driven Dynamic Routing of LLM Experts for Financial Trading” by
