The current standard for Conversational AI is entirely transactional. A user logs in, asks a question, gets an answer, and leaves. When they return the next day, the system treats them like a complete stranger.
As developers, we spend countless hours optimizing database schemas and crafting seamless state management for our web applications, yet we allow our AI integrations to remain completely stateless. We force users into a frustrating loop of constantly re-explaining their preferences, their past issues, and their account details.
The next major leap in UX isn't just about faster models or larger context windows—it is about moving from transactional AI to relational AI. Agents need persistent, decentralized state.
If you are ready to tackle this infrastructure challenge, the DeepSurge Hackathon Portal is currently hosting the Walrus Session 8: Chatbots That Remember developer challenge.
The Hackathon: Move Beyond the Blank Slate
Running from September 18 to October 9, this session challenges builders to eliminate the "blank slate" problem in AI by integrating decentralized memory.
By leveraging the Walrus Memory GitHub repository, developers can build or retrofit a chatbot that actively stores and recalls context across multiple sessions, devices, and days. The goal is to prove that when an AI retains long-term user state, the entire interaction paradigm changes.
You can apply this to almost any domain:
- Customer Support: A widget that remembers a user's open ticket from yesterday.
- Internal Tools: An HR or IT bot that knows an employee's department and past requests.
- Gaming: An NPC that holds grudges or remembers past alliances.
- Education: A tutor bot that tracks a student's struggles over an entire semester.
Submission Criteria for Builders
To qualify for the prize pool, you need to deliver a working implementation with proven real-world utility:
- Integration: You must use Walrus Memory to persist and recall context.
- Real-World Proof: The system must be tested by at least 3 distinct users, generating a minimum of 10 memories per user.
- Deployment: Your chatbot must live somewhere accessible—a web widget, a Discord server, Telegram, Slack, or a CLI tool.
- Open Source: A public GitHub repository with clear setup instructions is required.
- The Builder's Log: You must publish a 500–800 word technical write-up on Medium or Inkray detailing your build process. The judges want to see the exact moment your bot leveraged a past memory to improve the conversation, as well as any friction you experienced along the way.
The $2,500 Stablecoin Prize Pool
Whether you are a seasoned AI engineer or just starting to tinker with local models, there are multiple tracks to win:
- Best Overall Chatbot ($900 Total): Awarded to the top three projects that demonstrate the cleanest code, best documentation, and most impactful use of persistent state (1st: $500, 2nd: $250, 3rd: $150).
- Beyond the Big Two ($300 Total): A dedicated track for developers who step away from standard Anthropic or OpenAI APIs. Build with open-weight models (Llama, DeepSeek, Mistral) via Ollama/vLLM, or use alternative providers like Google Gemini or Groq to claim one of two $150 prizes.
- Best Technical Write-up ($300 Total): Three $100 prizes for the most honest, clear, and helpful articles documenting the developer journey.
- Bug Bounty ($500 Total): Find reproducible bugs or friction points in the Walrus Memory SDK? Log them on GitHub for a chance at five $100 bounties.
- Community Promotion ($500 Total): Five $100 prizes for developers who help expand the reach of the hackathon by sharing it in external developer communities and forums.
How to Enter
It is time to stop building bots that start from zero every single time.
- Register Now: Head to the DeepSurge Hackathon Portal to lock in your spot.
- Review the Docs: Dive into the Chatbot Reference Architecture to understand how Walrus Memory handles persistent state.
- Join the Hacker Community: Hop into the Walrus Discord to brainstorm, troubleshoot, and team up.
Submit your final project and article via the Official Airtable Form by October 9th.
Top comments (0)