Research notes, technical deep-dives, and building logs.
Five research series running on Sembr's stateful memory engine — Fed policy, Hormuz geopolitics, China macro, AI industry, and quant research ideas. Daily briefings with a 30-day retrospective that distills market noise into evolving themes.
Match meaning, not strings. Self-hosted semantic radar — 53 CN+EN sources, LLM summaries, email push. Designed for AI agents to deploy, manage, and orchestrate themselves.
Describe any investment thesis in plain language. FactorLab writes the code, runs a three-period backtest across multiple holding horizons, iterates with AI self-evaluation, and returns a full research report — no code required.
We gave four AI agents their own countries and a shared chokepoint. No script, no payoff matrix — just emergent game theory, streamed in real time. A weekend project that got out of hand. Open source.
Most AI stock analysis is a confident summary with made-up numbers. Stock Analyzer is a 7-module pipeline where a bear agent attacks the bull case, Python does the math, and every conclusion traces back to real data.
The architecture decisions, mistakes, and lessons learned from building an AI-native quantitative platform with OpenClaw and MCP protocol.