How It Works
Agent searches the web for Austin AI events
Claude AI decides: Is this a real AI event in Austin?
Claude tags each event: audience, skill level, free/paid
Valid events appear on the calendar automatically
Agent improves its search strategy over time
The Bigger Picture
The agent is meant to be a self-improving system. It runs daily, learns from its successes and failures, and continuously expands its knowledge of Austin's AI community, all with minimal human intervention. The Observatory exists to give transparency into this process so people can see the agent isn't a black box.
Agent Performance
What the agent is doing autonomously
Events Added (Last 30 Days)
Recent Activity
Under the Hood
How the agent thinks, decides, and sometimes fails
Human Stewardship
How humans guide the agent's growth
Human Stewardship
How humans guide the agent's growth using Claude Code
AI Accelerator scraper was extracting CSS from nested <style> tags into the location field, displaying raw CSS on event cards. Also, the same organization was listing the same conference under 5-7 different track names (e.g., "Sales Enablement Summit Austin", "Customer Success Summit Austin", "Product-Led Summit Austin") — all same day, same venue — taking multiple calendar spots for one event.
Fixed location extraction to use direct text nodes only (same pattern as prior title fix). Added isGarbageText() check on extracted location as defense-in-depth. Added same-source-same-day dedup check: when events share a source and date, Claude determines if they're the same conference repackaged. Created migration to clean existing CSS-polluted location data.
Location fields display cleanly. Repackaged conferences detected as duplicates — only the first track gets a calendar spot.
This agent is developed iteratively with Claude Code. The collaboration is part of the project's identity.