Pushing the frontier of memory and continual learning for agents. Add memory to your agents today
San Francisco, CA
Joined May 2024
- Tweets1.4K
- Following52
- Followers30.5K
- Likes1.8K
Pinned Tweet
what does supermemory actually do?
memory and context
for businesses and people
deployable anywhere
scalable and cheap
hey
many of you don't know what we actually do at @supermemory, so here's a short intro:
TL;DR is: if you're thinking about memory / context in any way, you should use supermemory.
- we are a team obsessed on building the best memory and continual learning system
- we primarily sell to companies building agents. Today we serve many thousands of companies (and tens of millions of users that they cater to)
- supermemory is very scalable and people use us because we're at the pareto of being good quality + being the most affordable way to do memory.
- you can also run supermemory fully locally, host it on your premise, or VPC (talk to us) teapawt.net/supermemory/status/206…
- you can also use supermemory for knowledgebases and retrieval, and we're better and cheaper on that front too. teapawt.net/supermemory/status/210…
- you can use supermemory with your own personal agents too (claude code, openclaw, etc.)
- we have a free plan so you can use it on your note taking sideproject :)
- it is that cheap and fast because we build the infra ourselves -- our own database primitive, and a model that is really good at long horizon learning
- it is different to other memory systems because we also think a lot about harness/product side. We're the only ones with hooks + tools approach, a user profile is created for everyone and injected into the model with hooks.
- you can tune everything about supermemory. you can customize it.
- think of it as a living, breathing database. teapawt.net/supermemory/status/211…
- but in reality, it's kinda like a graph teapawt.net/supermemory/status/211…
- our database can also be represented as a filesystem. teapawt.net/DhravyaShah/status/206…
- i have been working and thinking about this for 4 years now. the memory benchmarks are not very telling of what people need in the real world. I've spoken to 10,000+ people building in this space. a lot of things about supermemory's product are built with extreme care teapawt.net/supermemory/status/210…
- the infra: is really scalable. we have dreaming etc. which dynamically decides when and how to process information teapawt.net/supermemory/status/210…
- one more reason it's really cheap is because of a really banger model we built that is cheap but better than frontier teapawt.net/supermemory/status/209…
TL;DR: you should just use supermemory.
teapawt.net/supermemory/status/208…
supermemory retweeted
@livingbrain seems to be a new AI memory competitor, but I don't see a free tier to test out the product.
@supermemory has a free tier and it, along with their other tiers, seem more flexible.
Guess I know who I'm staying with.
supermemory retweeted
supermemory labs experiment:
What if you just virtualise the web as a filesystem?
it is MAGICAL
We built preprint, the live web as a filesystem for your agent
- every tab is a markdown file it can cat and grep
- to act, it appends one line, like click this link
- a real Chromium does it and the file rewrites
- a diff shows what changed
npm i -g @supermemory/preprint
We built preprint, the live web as a filesystem for your agent
- every tab is a markdown file it can cat and grep
- to act, it appends one line, like click this link
- a real Chromium does it and the file rewrites
- a diff shows what changed
npm i -g @supermemory/preprint
supermemory retweeted
Added @supermemory bot to my @bot and it's been fantastic so far. I'm so happy that this integration finally exists, I've been wanting it for a while. No issues, clean and seamless setup, and it works with @cursor_ai if you use it.
supermemory retweeted
PSA: case study unlocked 🔓
@supermemory had an agent benchmark Context.dev against their existing providers.
A few hours later, they were live in production.
We now power their web crawler, website extraction, and PDF extraction. One API.
“Never looked back." - @MaheshtheDev
Full case study below!
supermemory retweeted
we already do this in production!!!
as always, you should just use @supermemory
None of the 11 agent memory systems in a new paper beat plain RAG over the raw chat history on LongMemEval-s.
The paper is Memory Tax (arXiv 2610.34073). The baseline is basically the boring thing: chunk the raw history, hybrid search, a reranker, and let the model ask one follow-up query. With Qwen3.6-27B that scored 81.4. The best memory system, A-Mem, got 79.4.
It's not cheaper either. On their GPT-5.6 Luna run, A-Mem spent 117 hours building memory offline. RAG spent about 7 minutes indexing and scored 90.0 to A-Mem's 84.0.
The useful part is where the loss comes from. Every summarize or extract step is a chance to drop the fact you'll need later. In OpenViking, 38% of sampled raw turns had no memory unit at all. The rest is retrieval: the fact survives, but it gets split up or rephrased and the query stops matching it.
Their fix is small. Keep the raw log, retrieve from it alongside memory, and rerank both. It helped on all 3 systems they tested it on.
Fair caveat: the authors call LongMemEval-s retrieval-friendly, so memory has less room to help there. But if you're building agent memory, put RAG over the raw history in your eval table. If you can't beat it, the memory isn't adding anything yet.
Paper: arxiv.org/abs/2610.34073
supermemory retweeted
muse, dots, instinct... everyone is building a general assistant for everyone
but the real opportunity is the best agent for nurses, truckers, teachers, students...
so we’re running a race with @onkeiki @maritime_sh @supermemory and @pumpgtm
reach 1,000 agents and get up to $20k in llm credits and a trip to the dream city... (sf)
supermemory retweeted
the only console that your agents want in 2026
supermemory retweeted
you can get @supermemory to:
set-up, use, and fully manage supermemory
npx supermemory setup
pls use this and comment a screenshot below i will love you for life
with one command, now your agent can use supermemory and fully run it
use it
- add memories
- search them
- forget what's out of date
run it
- create API keys scoped to one namespace
- invite teammates
- see usage and logs
- connect tools like Notion
npx supermemory
use it today (for free)
console.supermemory.ai/
