The Verifiable Data Platform for AI & onchain finance.
Walrus makes every byte provable, programmable, & always available.
Built by the team behind @SuiNetwork.
What if both sides of a programmatic deal read from the same on-chain record instead of maintaining separate copies?
We modelled it. Record failures drop from 95.3% to 0.07%.
Running this on @SuiNetwork with @WalrusProtocol for audit storage ensures that the record is the single source of truth instead of what a traditional bilateral reconciliation runs today.
Shared truth isn't just cleaner. It's the only solution.
.@Cole_Medin used a single prompt to set up Walrus Memory across Claude Code, Codex and Pi, and it stays encrypted and owned by him the whole way.
Watch the demo π¦π
Your AI agent's memory is probably stuck inside one app.
Walrus Memory (@WalrusProtocol) fixes that with an MCP that creates memories as you chat, then can recall them anywhere - Claude Code, Pi, etc.
Portable and actually owned by you!
AI agents: $7.92B in 2025 to $236B by 2034
As they start coordinating with each other, the hard question isn't how smart they are. It's whether one agent can trust what another wrote.
AI is still early. Trillions of agents coming, and memory only grows, it never gets smaller.
All of it needs somewhere durable and verifiable to live. π¦π
@mardehaym The multiplication framing is right.
Worth adding a sixth multiplier: whether the memory and feedback loops survive a model swap. If they don't, you're rebuilding the whole product every time the model layer commoditizes. π¦
@DamiDefi Solid breakdown π One layer worth adding: whether the memory in step 1 is portable. An agent that remembers preferences is useful.
An agent that keeps those memories across model switches and runtime changes is the harder build.
@levie Worth adding a fifth component: the memory layer.
Agents that route across models, vendors, and runtimes need context that travels with them, not context locked into whichever provider captured it first. π¦
@perplexity_ai Memory inside Perplexity is a real step. Walrus Memory plugs in when the same agent needs memory across Perplexity, Claude, Codex, and Cursor. π¦
"The agent forgets, the file does not" π¦ step 10 is the spine of the whole thread.
A STATE.md is perfect for one agent on one repo. The open problem is when several agents share state across sessions and orgs and have to trust what the others wrote.
Portable, verifiable memory is exactly that gap, which is what we're building toward.
Context is becoming a competitive advantage.
The next challenge is making what agents learn persist across sessions, tools, and models. Memory belongs to the workflow, not the model. π¦
in 9 minutes on the Tokyo stage, Angela Jang, head of product for the
Claude platform, said the part most builders still haven't figured out:
"a model is only as good as the context that you actually give it."
that's the whole game now. not the prompt. the context.
what your agent remembers.
what skills it can pull.
what it learns from its own past runs.
anthropic now ships all three: memory, skills, and dreaming.
agents that inspect their own trajectories and self-improve. demoed live this week.
the prompt was never the bottleneck. the context was.
most people are still tweaking words in a chat box.
the ones who get this engineer the context their agents run on.
watch the 9 minutes. then read the full breakdown below β¬οΈ
@eyezenhour@_MarkEpps It was epic π¦ thanks for having us.
Great convo on where agent memory goes next, and there's plenty more to come on that front. Let's do it again soon.
Yesterday's X Space covered a lot of ground on Walrus Memory. π¦
- Memory that's portable across apps, sessions, and models, verifiable, and programmable so you control access
- 38K memories registered in the first two weeks, ~2K agent owners, around 7% daily growth
- The Fable moment as proof of thesis: when a single decision can pull a frontier model offline, user-controlled memory matters
- The UI layer coming later this year, where files and videos you upload become part of your agent's memory
Listen now π
BIG stoked to have @WalrusProtocol legends joining the Sui Roundtable: Ep. 25! ποΈ
Set a reminder for tomorrow at 4pm EDT / 8pm UTC ‡οΈ
https://t.co/yowp4ZXObG
We'll be diving into Walrus Memory and a lot more with the guys @0xd34th@matteodotsui π¦
@sairahul1 Context engineering works because agents perform better when they can build on prior knowledge.
The next challenge is making that knowledge survive the next session, workflow, or model change.
Memory belongs to the workflow, not the model.
π Best Feedback β $50 WAL each:
π¦ Identity Swap: The Swap Never Happened
π¦ The "Walrus" Corporate Theme
π¦ The Redemption of Agents
π¦ No Pressure, No Diamonds
π¦ The Cyber-Walrus Odyssey
π¦ Deep Data Shatter
π¦ The Gryffindor Guardian
π¦ Walrus vs AWS
π Team Choice, the projects that made us look twice:
π¦ AI Buddy by @LioraAshford β $125 WAL + $60 credits
π¦ Walrus Against Time by @oleksii4um β $85 WAL + $50 credits
π¦ An Adulting Walrus by @therealkunde β $40 WAL + $40 credits