# gg

- **Event:** [Built with Opus 4.7: a Claude Code hackathon](https://cerebralvalley.ai/e/built-with-4-7-hackathon)
- **When:** Apr 21 at 12:00 PM – Apr 27 at 2:00 AM (EDT)
- **Where:** Online
- **Team:** [Ray Tan](https://cerebralvalley.ai/u/Raytan)
- **GitHub:** https://github.com/101011101/Davaut/tree/ultra-mvp/davaut
- **Demo video:** https://www.youtube.com/watch?v=m1O1Yj3NHbY
- **Gallery:** https://cerebralvalley.ai/e/built-with-4-7-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/built-with-4-7-hackathon/hackathon/gallery/248

Match the chaotic brilliance of a human to the strategic, operational nature of AI.
                                                                                                                                                                                                        
  Davaut is a context-management substrate that ends two things at once: AI hallucination and the 100k-token context dump. Today's AI tools have the memory of a goldfish — you paste your life in, the 
  model wades through all of it to answer one focused question, you start fresh tomorrow, repeat. It's expensive, shallow, and forces you to be the memory system.                                      
                                                                                                                                                                                                        
  Davaut turns your work — docs, chats, code, plans, email — into a graph of nodes organized by layers (relations, tags, heat, importance, session timelines) treated as composable dimensions. A       
  workspace is a canvas. The Librarian, an Opus 4.7 Managed Agent, composes those layers like dimensions, assembles a course (a tight, mutable bundle of just the nodes that canvas needs), and serves
  it. Overnight, Dream — a second Managed Agent — consolidates: decays heat, promotes important tags into nodes, generates summaries. Your second brain gets smarter while you sleep.                   
                  
  The result is threefold. Cheaper — you stop burning tokens on irrelevance. Smarter — the model goes deeper because it isn't drowning. Grounded — the Librarian only returns nodes that exist, so the  
  AI on top physically cannot hallucinate. Every claim anchors to a real node with a real provenance trail.
                                                                                                                                                                                                        
  Davaut plugs into the SaaS you already use — Google Docs, Notion, email, your repo — no migration, originals sealed forever in the immutable Raw chunk. It's built for anyone running parallel        
  threads: the hacker juggling architecture and three half-formed ideas across linked canvases that inherit each other's context, and the founder who needs the ultimate secretary that pulls exactly
  the right context the moment it's needed.                                                                                                                                                             
                  
  Davaut is the memory under the model.

---

Markdown version of https://cerebralvalley.ai/e/built-with-4-7-hackathon/hackathon/gallery/248. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt