Hackathon: Build a Chatbot That Actually Remembers with Walrus Memory

Daunjuan — Mon, 21 Sep 2026 17:03:45 +0000

*Walrus Memory gives AI agents persistent, portable memory that continues beyond a single conversation and is not tied to one model or provider.
*

Instead of repeatedly passing entire chat histories back to an LLM, developers can use Walrus Memory to store the information that matters—preferences, decisions, previous problems and useful context—and recall it when it becomes relevant.

That addresses several problems that appear when a chatbot moves beyond the demo stage:

Returning users have to explain everything again
Previous troubleshooting steps and decisions are lost
Context gets buried inside long, expensive transcripts
Changing models or providers can mean starting over
A chatbot that appeared intelligent in one session behaves like a stranger in the next

We’re running an online hackathon to see what developers can build when their chatbot has memory that persists.

Build something new or add Walrus Memory to an existing chatbot. Use any model or runtime, deploy it somewhere people can actually interact with it, and show how its behaviour changes when it can remember previous conversations.

You can build with GPT, Claude, Gemini, Llama, Mistral, Qwen, Gemma, DeepSeek or another hosted or locally run model. There is also a dedicated prize category for projects primarily using models outside OpenAI and Anthropic.

The hackathon offers $2,500 in total rewards, including prizes for the best chatbots, open and alternative models, technical articles, and useful bug reports.

Submissions close on October 9, 2026.

The full description, rules, resources and registration are available here:

https://www.deepsurge.xyz/hackathons/c0141a4a-21be-4009-bc63-7c168608c849

DEV Community: Daunjuan

Hackathon: Build a Chatbot That Actually Remembers with Walrus Memory