I'm Daniel Kaufman. Kinetic Alpha started from a simple thesis — that excess return is generated through the intelligent movement of money — and grew into something broader: a working demonstration of what AI now makes possible in financial engineering. The projects here aren't the point. The point is that each one was built in days by one person, and that the same capacity is available to anyone who can describe what they want.

The four convictions

Excess return follows the movement of money. Capital that sits doesn't compound; capital that moves intelligently does. The framing isn't novel — what's novel is the speed at which a single person can now ship a defensible position on it. Every project on this site is, at root, an attempt to expose a structural inefficiency in how capital is priced, margined, or routed, and to make the exposure something you can actually drive.

Mental capital is the scarce input. The binding constraint on most investment processes isn't data, model quality, or compute — it's attention. A risk number that takes thirty minutes to compute well is a risk number that gets computed wrong, late, or not at all. The dashboards here are designed around the inverse principle: push routine analysis down to seconds so judgment can be spent on the parts that actually require judgment.

You are the developer. AI has collapsed the time-to-solution curve. The prediction-market margin framework is a 16-page paper, an 8-cluster Monte-Carlo model, and an interactive dashboard — built end-to-end in a single working week. The ICE energy decomposition library is 108 factors, 107 contracts, a Python engine, a parity-checked JavaScript engine, a verified reference table, and a portfolio dashboard — also built end-to-end, also in days. That capacity is now in your hands. Kinetic Alpha's most important job is to make this obvious.

Data ecosystems are the foundation. Every financial system is downstream of its data layer. The interesting work of the next decade isn't in model architecture — it's in building the open, verifiable, machine-readable data ecosystems that let those models be applied to real markets at all. Increasingly, that data layer lives on-chain: blockchains are uniquely good at capturing value in systems where provenance, auditability, and shared state are load-bearing. The next phase of this site will lean into that.

What you'll find here

• Live dashboards. Each project ships an interactive surface you can drive — change positions, sliders, conventions — and watch the numbers move. They are the whole point.
• Papers. Short, well-typeset documents for each project that spell out the assumptions, the math, and the choices other writeups tend to skip. If a model's convention isn't documented, it isn't a model.
• Code. Reference Python implementations with parity-checked JavaScript engines behind the dashboards. Tests and reference tables ship with everything. Drift between "what the paper says" and "what the dashboard computes" is a category of bug Kinetic Alpha doesn't tolerate.

What's next

The roadmap leans hard into the data-ecosystem thesis. On deck:

• An on-chain data layer for the existing dashboards — price marks, reference tables, and position data published in a form that any agent or app can consume.
• Cross-venue divergence scanning across prediction markets — Kalshi, Polymarket, Manifold — built on a human-curated question-link table so the comparisons are actually meaningful.
• A portfolio-level allocator that consumes the margin framework outputs as inputs, so the same engine that prices risk also shapes position sizing.

Contact

dkaufmanrisk@gmail.com. Always happy to compare notes on margining, clearing, on-chain finance, or building agent-native research workflows in any market where the textbook ends a few pages too early.