Marketing
- Kept X and Telegram active with real screening outputs, videos and macro regime reads while the stack was being rewired underneath — less noise, more signal.
- Updated the Messy Virgo GitHub org profile, repo descriptions, and roadmap to reflect what is actually live and what is being built now, so any developer or allocator landing on GitHub gets an accurate picture of where the project stands.
Business & Operations
- Confirmed the start of Phase 3 – Autonomous Trader through actual development: the fund agent stack now runs screening against a real paper trading fund (e.g. mvf-messybased) with defined universes of pre-screened Base tokens, producing auditable daily shortlists.
- Separated operations setup into discrete, reproducible scripts and introduced a secure test environment layer, so spinning up a full Messy stack is a documented, repeatable process rather than institutional knowledge.
Product Development
- Finished the token-screening data backbone and automated daily batch due diligence, now running productively for around 800 Base tokens in less than 2 hours.
- Redesigned the interface for our AI agents from an MCP server to a Messy Virgo CLI optimized for AI agents, and redesigned skills and agent definitions to support the full screening process.
- Improved the agentic screening workflow, including capabilities to run custom tasks sourced from both Messy-specific and external resources.
- Extended existing skills so AI agents can work with historical screen runs, querying more than 120 historic KPI and Score values across tokens for screening analyses and back-testing.