Dhravya Shah
@DhravyaShah
19. Chief builder, Solo Founder, CEO @SupermemoryAI. "extraordinary" @O1Visa. Lifelong learner and serial shipper. contributing to AGI with memory
We just solved the context length issue 🤯 Infinite Chat with llms ⚡ Yes, that's right. one single line to extend the context length of any llm. by millions, and even billions of tokens.
Introducing: Infinite Chat API 💥 Extend the context length of ANY model ...while saving 90% of your tokens and cost AND improving performance! One single line to switch. Available to use now
i ran every possible scenario. in none of them i get to $100M ARR without becoming this.
why do you have to be a fucking influencer to be a great founder 😭😭😭
composio is fire
Agents aren’t reliable. They don’t learn from experience. At @composiohq, we provide skills that evolve with your agents @lightspeedvp gave us $25M to make agents usable
we went to paying $100+ a day to OpenAI to $0 👀 cc @googleaistudio
supermemory connectors tutorial ->
Did you know? supermemory has built-in connectors to Google, Notion, Onedrive and others, which means it's really easy to sync data to and from these services! supermemory.ai/blog/building-…
we just need one more final standard bro, please bro, just one more standard. please bro bro, one more standard. how will agents communicate? standards. one more standard bro, please bro
one standard to let agents discover each other regardless of standards sounds familiar
i genuinely love every single thing I do at supermemory we are solving some really hard problems and feels like every morning i'm just extremely pumped about what's going on we don't make big money yet but very soon, i know i will win.
what's some breakthrough business models you've seen? Some that come top of my mind are - Cloudflare - Huggingface what else?
The secret to being good at stuff is to take personal offense when anyone is better than you
give your employees AI credits they said it will be very cool they said @CodeWShreyans out here contributing to cursors 1B ARR pulling an all nighter for an upcoming launch
Walked into the PlanetScale office, did a (very) complex migration for ~6 hours, without ANY downtime whatsoever. Walked out with a faster, better, and cheaper supermemory Very excited for this because our customers fully benefit from this
We partnered with @PlanetScale to give you the most performant memory engine engine you've ever seen This means, for our customers supermemory is, - 40% faster - Cheaper - More reliable and performant (up to 1000 QPS!) supermemory.ai/blog/supermemo…
This is actually possible, using sparse auto-encoders A lot of Anthropic's interpretability research has led to this technique where if you blow up the embeddings, you can extract "features" for a given input read more: thesephist.com/posts/prism/
We need a reverse embedding API remember word2vec's King – Man + Woman = Queen? How can we do that with LLMs? LLM providers should offer an API that generates text from a given feature vector. Then do: RevEm(Em(King) - Em(Man) + Em(Woman)) This would open up a whole new…
feeling very, very extremely e/acc we're at one of the inflection points of AI everyone in the world has the opportunity to push the slope forward and there's so many more to come - robotics, quantum, genomics, energy, spaceflight,... so many. what a time to be alive
the future is today
2024: Voice Cloning 2025: What about personality cloning? Hume’s voice AI can now not only mimic your voice but also speaking style and language. It’s now available via our TTS and new speech-to-speech model, EVI 3, which is also launching today.
insane!!
Lovable just raised $200M at a $1.8B valuation led by Accel. This all started unexpectedly with me calling my friend at 6AM to go for a walk. I've never shared this story before: (thread)
customers are loving graphs inside @supermemoryai You can now view how the supermemories of your users evolve, as documents and memories are added to it, with: temporal understanding, forgetfulness relevancy bias (..& more) literally just postgres btw. no graphRAG bullshit
there's beauty in simple things in life
BEHOLD: Supermemory's graph You can now view the entire constructed supermemory graph on the supermemory console. You can view the documents, memories extracted out, and more information about the links between the memories, and how it's self-forming and self-evolving ✨