Pushing the frontier of memory and continual learning for agents. Add memory to your agents today
San Francisco, CA
- Tweets 1,332
- Following 51
- Followers 28,548
- Likes 1,558
We're open sourcing our company brain, a multi-player harness that acts as a teammate in a slack,
with frontier memory! Link below, you can set it up and self-host in 5 minutes with one click.
We're open sourcing the company brain, a multi-player slack harness that we worked on for months, it's an AI employee with all the knowledge of your company.
You can deploy it with one click in minutes.
I go over the full architecture, in this blog!
11
29
397
38,837
Usual memory systems learn the moment data flows in, then never look at it again.
@supermemory 𝚍𝚛𝚎𝚊𝚖𝚜.
Status done means the chunks are indexed. Memories come from a second phase, where related documents are grouped so memories form from coherent units, never one isolated write.
When you go quiet, or enough new context has piled up, supermemory enters a dream cycle on its own. During a dream it:
- merges fragments that belong together
- reweights old facts against everything since
- resolves contradictions
- derives facts you never stated in one place, each traceable to its sources
The output is new memories, a graph of derivations between them, and a picture of the user that updates as the evidence changes.
If you use supermemory in any way, this already works. Dream away.
4
2
36
8,622
@supermemory console updates, shipped this week!
- two-factor authentication, enforceable across your whole org on scale and enterprise
- the memory graph renders while it loads, explore connections before the last document is in
- a rebuilt sidebar for data, developer tools and settings
- disconnecting a connector now asks whether its documents leave with it
3
2
19
5,565
We reverse-engineered Instinct's memory system and implemented it with supermemory's architecture
I spent 3 days reverse-engineering Instinct's memory. It's beautifully simple, and works extremely well.
i also wrote about how you can implement the same in supermemory in 60 lines of code!
4
17
428
54,527
supermemory retweeted
markdown vs @supermemory
in markdown only memory systems, the agent may have to look through and traverse files to answer things. This is wasteful and slow! The files also get stale really fast, and there's no forgetfulness, updates, etc. The agent has to just look through things.
supermemory is kinda like a dense interconnected markdown system, except that we automatically construct almost an 'instant' file that will directly one shot what the agent needs. That context has the whole history of the thing, removes forgotten info, and even related things that the model should probably know.
This information is dense, fast, and dynamically constructed, so you have a living breathing improving system.
So yes, you should use supermemory instead of markdown
11
3
86
6,228
supermemory retweeted
every day I'm even more bullish on how beautifully simple and composable supermemory's architecture is. And this is with incremental improvements throughout years of being with customers and really really caring about THEIR products.
@supermemory will be the default way to do memory in the agents era. There's nothing else that makes sense to me.
7
4
65
5,543
a bad memory is a pile of chat turns ranked by similarity. the fact from march sits next to the one from today and both look equally true.
so what is a good memory?
- scoped to a user, project, tenant, or any entity, with reference material kept as documents in that same scope
- written when it's worth remembering, so casual chatter is less likely to stick
- kept as a fact that holds until updated, a preference that strengthens with repetition, or an episode that fades unless it mattered
- related with updates, extends, or derives, so the latest fact wins, history stays, and a guess ranks lower until confirmed
- atomic, time-aware, and willing to forget when the window closes
- attributed to source documents and traceable
That is supermemory.
4
2
27
2,793
Introducing the supermemory plugin for @AIatMeta muse code.
Shared (super) memory capabilities that feels like magic!
Releasing Muse Code in beta today. It's a terminal coding agent that takes on complete software engineering tasks across large repos: planning changes, writing code, validating the results. Powered by Muse Spark 1.2, a coding-focused model update.
5
2
33
9,204
In personal assistant and chat based agents, 𝚑𝚢𝚋𝚛𝚒𝚍 mode in supermemory can greatly improve your agents performance. Here's why:
Typically in memory systems, the main model, or another model has to figure out what's important to learn - we do this too! But, it's impossible to be able to accurately predict everything important for learning.
In supermemory, learnings are formed for things that are important for future context and needs to stay fresh, or forgotten. But, raw context is also indexed.
If the memories don't include all the information that the model needs, supermemory will automatically add parts of the previous conversations to enrich the context!
In the background, learnings are made with this info, now that supermemory knows that it has importance.
Fun fact: In supermemory we not only learn memories, but automatically index the conversations as well.
Below is an example of how @supermemory works in the Vercel @eve harness.
everything works together in a harmony. Memories are formed in a coherent manner (even for things never said in a single message are learnt naturally), and the model doesn't have to choose to "remember" things. Updates, relations and forgetfulness is encoded in the data structure directly.
It's multi-modal too! The context is then injected back into the model, for magically personalized and fast responses.
1
19
2,880
Your agents are interchangeable but your context shouldn't be.
@supermemory is the persistent context layer across all your agents
Swap the model and keep the memory!
7
9
157
21,932
Introducing the new supermemory console.
Revamped design, better UX, new functionalities!
10
11
232
46,651
Here's the details about why you should use it!
We built a native @supermemory integration for Eve, and it's available for everyone today!
We always put a lot of thought into the harness specific considerations for these plugins. So we worked hard and are co-lauching as one of the first memory providers available on eve.
What makes it good?
Supermemory automatically indexes the sessions and multimodal input, also learns things and keeps knowledge fresh, injects a profile, handles forgetfulness, and inject it into Eve's context
The model does not decide to "capture" or maintain memory. It is handled completely externally. So this is especially good if you are building something that needs to feel magically personalized, instead of tool calls firing off every prompt.
If the agent needs to, it can look up it's past sessions too, just like Claude / Hermes do. But, most things are implicitly provided to the model, without it asking for it (we think it's important!)
it also uses profiles to optimize the context/provide info and injects it at the start of the conversation.
We inject 7x less tokens than other providers, because our model is learning and handling that context too.
This leads to tokens saved (main model doesnt have to do the traversal again and again), better experience, more features (session search, multimodal memory etc.) in a few lines of code.
here's a little graph of how it works.
2
277
Introducing native support for supermemory in eve
It supports scoping, learning, and a lot of harness-specific optimizations for eve.
we use hooks+tools to implicitly add memory to eve agents
It's one of the first memory integrations. Few lines to install
eve agents now have memory.
Use built-in file memory, a memory service, or your own provider to retain context across sessions.
𝚎𝚟𝚎 𝚊𝚍𝚍 𝚖𝚎𝚖𝚘𝚛𝚢/𝚏𝚒𝚕𝚎
4
5
45
5,161
You can use supermemory to power your agents, today: supermemory.ai
2
27
7,428