Research & Experiments
How the agentic systems actually work.
Audits of real repositories and analysis of the methodologies behind them. Everything is reproducible: audits cite the repository, experiments publish their raw data, and live numbers link to where they come from.
Experiments
1 totalPre-registered means the design is published and locked but the run has not started. Running means data is being collected. Published means results are out, together with the raw data.
Articles
5 totalInfrastructureBeyond market data: what an AI investing agent needs to know about youResearch, portfolio access and order approval solve different problems. A comparison of the tools that carry an investor’s holdings, rules and decision history—and where Investboard fits.Oct 11, 20269 minMethodologyFrom TradingAgents analysis to a reviewable investment decisionWhat investor context changes in the research workflow, what the connector records, and how to test it without confusing a mandate check with an order block.Oct 11, 20269 minOpen Source10 AI investing repos worth watching right nowNot the ten biggest. The ten where the next six months will tell you something: fastest growers, execution engines going AI-native, and the infrastructure that agents are converging on. Live numbers, updated daily.Oct 9, 20264 minInfrastructureMCP is becoming the missing layer between AI and marketsBrokers and data vendors are shipping official Model Context Protocol servers. A survey of what exists, how the security model works, and why the interesting question has moved from access to judgement.Oct 9, 20265 minResearchThe rise of multi-agent investing systemsAnalyst teams, bull-versus-bear debates, risk committees and fund managers, all played by language models. How the architecture spread from one paper to the most-starred projects in the category, and what remains unproven.Oct 9, 20265 min