I built Wings Studio, an AI workspace that combines two patterns builders are excited about right now: agent gateways that can act through tools and messages, and LLM-maintained memory that compounds over time.
Instead of storing memory only as chat history, vector chunks, or Markdown wiki pages, Wings turns files, notes, tasks, decisions, meetings, and agent runs into typed objects in an inspectable graph. The demo shows the architecture of the system, walks through how the graph is built, and shows a Data Science Agent extracting entities and relationships from messy files.
