Journal
Posts and long-form series on AI, startups, venture capital, and more.
All Posts
Bankroll Management with Keeks: Dynamic Strategies
Dynamic Bankroll Management adapts bet sizing based on recent performance, volatility, and changing conditions. Learn the flexible approach with Keeks.
The Case for a Dumb Idea App
Ideas aren't tasks. They don't need due dates, priorities, or kanban boards. They need fast capture, a first step, and a path to execution.
python-skills: Packaging Library Development as Claude Skills
I turned my Python library development guide into a set of independently installable Claude skills. Here's what they are and how skill marketplaces work.
Training a Deep Q-Network for Trading
How we're using Deep Q-Networks to learn trading decisions, and the techniques that make training stable enough to actually work.
Bankroll Management with Keeks: CPPI
CPPI protects a floor value of your bankroll while betting a multiple of your cushion above it. Learn how to implement portfolio insurance with Keeks.
Ruff vs Traditional Bandit - A Performance Comparison
Ruff is 25x faster than traditional Bandit. Here's what that performance difference means for your workflow and when to choose each tool.
Reviving Graduate School MATLAB with Claude Code
How I used AI coding assistants to transform decade-old MATLAB research code into a modern Rust library with Python bindings. The origin story of ParGA.
Bankroll Management with Keeks: Fixed Fraction
The Fixed Fraction strategy bets a constant percentage of your bankroll on each wager. Learn when this simple approach beats more complex methods.
The Scientific Method as a Decision Framework
Most structured decision frameworks are really just the scientific method adapted to different time scales and contexts.
How I Write Software With AI
My workflow for building data-intensive services has gotten messier and faster. AI tools have made throwaway code cheap, which changes everything.