Thoughts on AI, systems design, and the craft of building software.
Weekly-ish dispatches on AI engineering and systems design.
AI coding agents burn 20–40% of their context window on inefficient codebase exploration. A hierarchical documentation system cuts orientation from 15–20 tool calls to just 1–3.
Announcing Stoneforge — an open-source platform that orchestrates multiple AI coding agents in parallel using git worktrees, automated task dispatch, and merge reviews.
Open-source passion is fading as contributions become career stepping stones and corporate talent-scouting tools. A case for systemic change to sustain the ecosystem.
Introducing TensorTrade, an open-source Python framework for training, evaluating, and deploying robust trading agents using deep reinforcement learning.
Applying feature engineering, Bayesian optimization, and the Sortino ratio reward to push average agent profit to nearly 850% on test data.
Experimenting with deep reinforcement learning and OpenAI Gym to build cryptocurrency trading agents that learn profitable strategies on historical BTC data.
Extending a custom stock trading Gym environment with rich, annotated Matplotlib visualizations that update on each step.
Building a custom OpenAI Gym environment for simulating stock trades on historical price data using reinforcement learning.