Shared, verifiable memory for connected AI agents | ChatGPT, Claude, Codex & more. Trusted by 1000s of apps 🌐 CA: 0xfC5A5c521C80d8e1d6B5Aee6454B8C9483C91BFf
Robinhood Chain
Joined October 2021
- Tweets3.9K
- Following295
- Followers24.9K
- Likes7.1K
Pinned Tweet
One memory. Every agent.
Lighthouse gives AI agents a persistent, encrypted brain that travels with you across ChatGPT, Claude and the rest.
Saved once, owned by you, verifiable on decentralized storage.
Lighthouse ($LLL) is now live @RobinhoodCrypto via @virtuals_io
{Link in comments}
Lighthouse retweeted
As intelligence becomes abundant, the scarce resource won't be intelligence itself.
It will be context.
who owns it, who controls it, and which agents can access what
i think memory will become one of the AI's most important commanding heights.
The benefits of abundant intelligence will accrue far outside of San Fransisco.
Spent the last year writing about capital, sovereignty, and the coming reordering of economic power.
THE COMMANDING HEIGHTS
Commanding-Heights.com
Lighthouse retweeted
Thankful to our users for trusting us.
Making something people want at @LighthouseWeb3
Half the work is making a decision. The other half is finding where you wrote it down.
This week, we launched Lighthouse Brains to help bring those scattered notes, docs and memories together.
Here’s what we shipped and explored, 25–31 October🧵
8/8
The part worth keeping is often the reasoning: why you chose this approach, what you already tried, what changed.
That’s the context you want waiting for you when you return to a project.
brains.lighthouse.storage
Some memories are too spooky to forget. 👻
Your AI shouldn't forget them either.
Happy Halloween from Lighthouse Memory. 🎃
Lighthouse retweeted
Just love these Lighthouse Brains as a new medium that I use to visualise memory, docs, connections and much more.
- works across private and public memories
- talk to the brain to search through memories stored
- with verifiable proofs, timestamp and more !
Better models can search history more effectively. They still need access to it.
The finding worth sitting with here: a stronger model cannot recover details that an earlier memory system left out. A small experiment, but a useful reason to test what your agent actually needs.
Alongside recall quality, there’s ownership and portability. Can you carry your context into the next tool?
That’s where Lighthouse fits: user-owned, verifiable memory that lives beyond any single model or session.
// MemoryTax: The Bitter Lesson of Agent Memory //
Great insights for agent builders; bookmark it.
As models increase in capability, they also improve at things like memory generation.
Agent memory frameworks are everywhere right now. This study measured whether they still help once the model can search raw history itself.
The results suggest that a strong model needs less designed memory and
better tools for searching the raw record.
On one 419-turn LoCoMo conversation, Mem0 beats plain full context only with GPT-4o-mini. Building Mem0's memory costs more than an entire full-context rollout.
With Claude Opus 5.5, the Pi coding agent searching the raw history scores 99% for under half the cost of full context. Its one miss comes from a wrong answer key.
Memory also has a ceiling set by the model that wrote it. A stronger reader cannot recover what a weaker writer left out.
The author clearly defines the scope: one conversation and one run per cell. It is still a cheap experiment to repeat on your own agent before adding a memory layer.
arxiv.org/abs/2610.29823
Chat with Paper: academy.dair.ai/papers/memor…
A vault door on a straw hut is a pretty good security review.
That’s the image in @VitalikButerin’s Ethereum Community Conference slide: formally verified contracts behind a frontend that can be replaced if someone hacks the server.
His slide points to static HTML pages on @IPFS as part of the solution.
Then look at Polymarket’s 29 October incident. A third-party provider was compromised, and researchers found a monitoring script designed to access trading sessions and attempt to move assets.
The code meant to watch the page had become a threat to the people using it.
IPFS helps applications check content against an expected identifier. But those checks must cover the dependencies too. An unchanged page can still load a changed external script.
At Lighthouse, this is relevant to both storage and memory.
Lighthouse Storage helps builders store and retrieve content through IPFS .
Lighthouse Memory gives agents context they can carry between sessions and compatible tools.
When that context starts informing trades, payments or enterprise actions, its origin and permissions matter as much as its usefulness.
Your agent remembering an instruction is convenient. Knowing who put it there is essential.
You remember making the call.
You just can’t remember which chat it’s buried in.
We put the Lighthouse Memory docs into a Brain and asked it how recall finds relevant memories.
It pulled the memories that answer it and lit them up in the graph.
Your project’s reasoning, open to explore and ask 🧵
16 public. 2 private.
Public memories are stored on Lighthouse, each with a CID anyone can check.
Private ones stay locked and show in the graph as a lock, so people see something is there without reading it.
Only your wallet can add to or change your Brain.
Brains run on the same storage 34,000 builders already use, with 50 TiB stored so far.
Open the Lighthouse Memory Brain, ask “How does recall find relevant memories?” and click the top result. Then publish your own.
brains.lighthouse.storage