# Ghost Browser

- **Event:** [WeaveHacks 3: Self-Improving Agents Hackathon with Weights & Biases](https://cerebralvalley.ai/e/weave-hacks-3-self-improving-agents-hackathon-with-weights-and-biases-7014fe80)
- **When:** Jan 31 at 9:00 AM – Feb 1 at 5:00 PM (PST)
- **Where:** 400 California St, San Francisco, CA
- **Team:** [Ashutosh Kumawat](https://cerebralvalley.ai/u/ashu), [MAHIMA ADVILKAR](https://cerebralvalley.ai/u/MahimaAdvilkar)
- **GitHub:** https://github.com/DunngenMaster/weaveit
- **Demo video:** https://youtu.be/7yCclyWmJ-A
- **Gallery:** https://cerebralvalley.ai/e/weave-hacks-3-self-improving-agents-hackathon-with-weights-and-biases-7014fe80/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/weave-hacks-3-self-improving-agents-hackathon-with-weights-and-biases-7014fe80/hackathon/gallery/36

We built a self-improving AI memory backend that prevents “fresh start” chats by continuously capturing a user’s goals, preferences, progress, and browsing context, then distilling it into clean long-term memories and reusable “what worked” patterns. When the user starts a new chat (ChatGPT/Gemini/Claude) or the conversation stalls, the system generates a compact context snapshot file and attaches it so the next model can continue instantly—while filtering unsafe requests, deduplicating repeated attempts, and only learning from validated successful outcomes.

---

Markdown version of https://cerebralvalley.ai/e/weave-hacks-3-self-improving-agents-hackathon-with-weights-and-biases-7014fe80/hackathon/gallery/36. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt