Tutorial seriesPublished2025
AI Agents in Python
A three-part progression from simple decision agents to LLM-powered agents and clarifying loops that ask before acting.
Systems. Code. Markets. Time.
Software, trading infrastructure, AI experiments, and unusual research.
Built and documented here.
Four pillars
Browse the site by interest, not just chronology.
The blog still matters, but the structure now starts with the main threads underneath it: projects, technology, trading, and time.
Work
Projects
Libraries, tools, browser extensions, trading infrastructure, and experiments that turned into real systems.
From goeph and backtesting tools to older frameworks like Nirman and Livebase.
ExploreCode
Technology
AI agents, Go systems, JavaScript, architecture, debugging, and the practical side of building software.
Current writing leans heavily into AI workflows, software design, and developer tooling.
ExploreMarkets
Trading
Backtesting, strategy evaluation, trading psychology, and the systems thinking needed to survive real markets.
A mix of code-first research, practical trading process, and uncomfortable lessons.
ExploreResearch
Time
Time cycles, ephemeris-driven research, celestial data, and the thread tying outcomes, astronomy, and pattern-seeking together.
Where celestial data meets quantitative methods and pattern research.
ExploreSelected work
A few projects that show how I think and build.
Go UI frameworkActive2026
godom
A Go framework for local apps where the Go process owns the DOM and the browser is just a screen. No Electron, no complex JS frontend, no API layer; go build gives you one binary that opens a tab.
Trading infrastructureActive2025
Go Backtester
A custom backtesting engine built for strategies that stopped fitting neatly inside off-the-shelf Python frameworks.
Go libraryActive2026
goeph
A Go ephemeris toolkit built on NASA JPL ephemeris data, born from porting a real Skyfield-based workflow. Now the bedrock for further tooling in this space.
One-off toolExperiment2025
SEO Analysis Tool
A sitemap-driven SEO analysis workflow built almost entirely with AI agents, then corrected by hand where the agents drifted.
Recent writing
Latest from the workshop.
Jun 9, 2026
What an AI agent actually is, and how to build one
What an AI agent actually is, how it differs from a plain LLM, and what I learned building several in Python.
Jun 9, 2026
How WebSockets actually work, and what bites you
How WebSockets work: the handshake, the frames, and the gotchas, learned from running real systems over them.
Jun 9, 2026
Prompt engineering examples that actually work
Concrete prompt engineering examples and the patterns behind them, drawn from how I use AI for code, research, and writing.
Jun 9, 2026
Claude Code: what it is and how I actually use it
Claude Code is Anthropic's agentic coding tool. Here's what it is, how it works, and how I actually use it across real Go and Python projects.
May 30, 2026
Backtesting the four classical harmonic patterns
Gartley, Bat, Butterfly, Crab on ES from 2020 to 2024. One looked tradeable; the cross-instrument check killed the universal claim.
May 29, 2026
I backtested six classical candlestick patterns. One held up.
Six candlestick patterns tested on NQ from 2020 to 2024. One held up. Two failed cleanly. One fired too rarely. Two were NQ-only artifacts.
Topic trails
Prefer a narrower thread? Start here.
AI & Agents
Agent workflows, prompting, validation loops, and where AI helps or burns.
Trading & Backtesting
Strategy testing, execution logic, risk, and practical trading process.
JavaScript & Web
Web development, browser tools, static sites, and older JavaScript work.
Go & Systems
Performance-minded tooling, compiled workflows, and system-level implementation.
Python
ML, analysis, scripting, and the places where Python still fits best.
Insights
Meta writing about work, tools, migration decisions, and how I think through tradeoffs.