Most chatbots suffer from the goldfish problem—they forget who you are the second you close the tab or start a new session.
The team at Walrus is hosting a dev challenge called Session 8: Chatbots That Remember, and they are looking for builders to fix this. The goal is to build (or retrofit) an AI chatbot that utilizes persistent memory across different sessions, users, and devices, and showcase what new experiences this unlocks.
The Details:
What to Build: A chatbot that retains context. It can be for customer support, onboarding, a game NPC, a tutor, or a companion.
The Tech: You'll use Walrus Memory to store and recall the context on mainnet.
The Prizes: There is a $2,500 prize pool spread across a few categories, including Best Chatbot, Best Article, and a special "Beyond the Big Two" track if you use open-source or alternative models instead of just OpenAI/Claude. They also have bug bounties if you just want to break things.
Deadline: October 9th, 2026.
If you are already tinkering with LLMs, LocalLLaMA, or AI agents, this is a great excuse to experiment with agentic memory and get rewarded for it.
You can find the full rules, the memory repo, and the submission portal here:
https://www.deepsurge.xyz/hackathons/c0141a4a-21be-4009-bc63-7c168608c849
Top comments (0)