Google’s Former Leaders Back Supermemory  ## Metadata - Author: [[Hard Fork AI]] - Full Title: Google’s Former Leaders Back Supermemory - Category: #podcasts - URL: https://share.snipd.com/episode/0ef70558-8c7a-4fb8-a6c1-fdec2f8a3e3d ## Document Notes ## Highlights - **Cross-Model Memory Breaks Platform Moats** - Super Memory is an API that provides persistent memory across multiple AI models and apps. - It dissolves single-platform memory moats by centralizing context across models and services. Transcript: Jaeden Schafer So this is an interesting company. In fact, this is one that personally, I believe we might actually start using over at AIbox because essentially what it does is it's like memory on ChatGPT where it takes everything You've ever said to ChatGPT and it stores it. And so when you talk to ChatGPT it remembers that, oh, a month ago you talked about going to this place or that this was your favorite coffee shop or that, you know, this is the specific Project you're working on and it will bring up information that you tell it. So it's, you know, ChatGPT is incredibly useful for this. Now what happens is there's thousands of companies out there with chatbots and different AI tools that don't have memory installed because it's kind of a bit of a complex topic. And so super memory is essentially an API that does that for companies. And the cool thing is that they do it across different apps. ([Time 0:01:10](https://share.snipd.com/snip/a3e9d72b-354e-4573-a268-d76d124778bd)) - **Prevent Vendor Lock-In With Shared Memory** - Integrate a shared memory layer to avoid locking users into one AI model. - Offer cross-model memory to make switching models seamless for users. Transcript: Jaeden Schafer So in the case of AI Box, for example, where I let people use 40 different apps, instead of just having all of your conversations siloed by conversations between Claude or ChatGPT or Google Gemini, integrating something like Super Memory would all of a sudden make it so if you're on AI Box and you are talking to any of the models, it has memory from anything you've Talked about. So it's kind of amazing because it makes it so that instead of, you know, just getting all your memory stuck in ChatGPT, which that was kind of the moat, I think, when ChatGPT came out with Their memory feature. They're like, oh, this is a good moat. No one's going to want to switch to Claude or Gemini if they have all of this memory, all of this, you know, good data built inside of ChatGPT. They don't want to give up these great chats. So as soon as you have a tool like SuperM, all of a sudden, there's no reason to stick with chat GPT, that moat goes away. ([Time 0:02:00](https://share.snipd.com/snip/a32f57c5-b666-4b1d-ae03-ae499fe27c4a)) - **Founder Built It During A 40-Week Challenge** - Founder Dravya Shah built Super Memory after a 40-week challenge of building something new each week. - He previously sold a bot, moved to the US, attended ASU, and turned one weekly project into a full product. Transcript: Jaeden Schafer Their founder is 19 years old. So very young. His name is Darva Shah. And it's interesting. He's originally from Mumbai. And this wasn't a company that he just like set off built and was like, this is going to be my thing. He actually started building this. Well, he formerly sold a company that he had. He had a bot because he's, he's done a lot of consumer facing bots in the last few years. And he sold one of those. It basically formatted tweets into nice looking screenshots for social media. He sold it to Hype Fury. And because of that, he earned enough money that he decided he was going to move to the US. Originally, he was studying to go to the Indian Institute of Technology. He moved to the US and he started attending ASU, very close to where AI Box is headquartered. So after he moved, he decided that he wanted to do a challenge. I've actually heard this from a bunch of other ASU students that I've talked to. So maybe this is a trend over there, or maybe this is kind of happening everywhere. But every week for 40 weeks, he was going to build something new. And one of those weeks he built super memory. So initially he called it any context, he put it up on GitHub. And back then it basically just let you chat with your Twitter bookmarks. So it would look at all the context, anything inside your bookmark, you could chat with your bookmarks and like, Hey, what are the top bookmarks I have about marketing, whatever, right? So you're chatting with it. But what he didn't realize was this is actually an incredible tool and use case for a much bigger problem than just chatting with your Twitter bookmarks for these AI models. And it's kind of interesting how he actually came to this conclusion. He did an internship at Cloudflare last year, and he worked on AI infrastructure there. And then he was also their developer relations lead. And so while he was there, he had some advisors, specifically the CTO of Cloud4Layer, which was Dane Knet, who talked to him about all those projects when he heard about SuperMemory. He's like, oh my gosh, please turn this into a real product because that's actually super useful. And so then this year he decided to work on it full time. ([Time 0:03:18](https://share.snipd.com/snip/c2734c38-371d-4423-b319-51864b9841de)) - **Cloudflare Internship Sparked Productization** - Dravya interned at Cloudflare working on AI infrastructure and developer relations. - Cloudflare's CTO encouraged him to productize Super Memory after seeing its potential. Transcript: Jaeden Schafer He did an internship at Cloudflare last year, and he worked on AI infrastructure there. And then he was also their developer relations lead. And so while he was there, he had some advisors, specifically the CTO of Cloud4Layer, which was Dane Knet, who talked to him about all those projects when he heard about SuperMemory. He's like, oh my gosh, please turn this into a real product because that's actually super useful. And so then this year he decided to work on it full time. And what's really impressive to me about this is that this is an API. It's not just like he built super memory for his platform that has, you know, lets you chat. ([Time 0:04:52](https://share.snipd.com/snip/2b76a1ab-2ad2-466f-ade3-5bb0ea3fe2cd)) - **Knowledge Graph Powers Personalized Context** - Super Memory constructs a knowledge graph from users' multimodal data to provide personalized context. - It supports documents, chats, emails, projects, PDFs, and assets to surface relevant context quickly. Transcript: Jaeden Schafer It's not just like he built super memory for his platform that has, you know, lets you chat. It's an API. And so any apps can plug this in, which is really interesting. Um, it basically builds a knowledge graph based on the data that you give it. And then it personalizes the context for the users. So it can support querying across months old entries for a writing or journaling app or searching for an email app. And then as the solution allows for multimodal inputs, it also can allow, for example, so like it's not just for text. It could do like, you know, it could help a video editor to fetch a relevant asset from a library. It can do documents, chat, projects, emails, PDFs. So it's really cool, right? It's got all of these different, all of this different data that you can stick in there, and it will remember them across the whole platform. So beyond just remembering what, what chats you have, you know, the types of things you've talked about, it can go and like scan through your, you know, your, your links, your Google Drive, your OneDrive, your Notion. There's also a Chrome extension that lets you add notes from a website. So this is really cool. And basically you build these knowledge bases one way or another, and this thing can plug into them when you're talking to AI models. It knows a lot about you. ([Time 0:05:24](https://share.snipd.com/snip/ab624d60-b739-430a-9500-fe8012b06858)) - **High-Profile Backing Validates Vision** - Super Memory raised $2.6M led by Sousa Ventures and Broder Capital with notable AI figures investing. - Backers include Cloudflare's CTO, Google AI chief Jeff Dean, and DeepMind product manager Logan Kilpatrick. Transcript: Jaeden Schafer I think that's really cool. So they raised $2.6 million. That was led by Sousa Ventures and Broder Capital. And there's a bunch of different investors. Cloudflare's CTO, who is kind of his mentor, got in. Google AI chief Jeff Dean. DeepMind product manager Logan Kilpatrick, a friend of the podcast, also invested in this. And so, yeah, overall, I think this is a really, really cool project. Apparently, Y Combinator contacted him, but it didn't work with timing ([Time 0:06:56](https://share.snipd.com/snip/0d8659a0-aea3-4caf-94d1-2a62d35067c3)) - **Move Fast To Win Investor Interest** - Move fast and ship to attract investor interest. - Rapid iteration and getting products into users' hands matters more than just having ideas. Transcript: Jaeden Schafer He said, I connected with him on X and what struck me was how quickly he moves and builds things and that prompted me to invest in him. I think this is one of the number one things that investors are looking for is how fast can you iterate on an idea beyond just having an idea? How fast can you actually get it into the hands of consumers? ([Time 0:07:54](https://share.snipd.com/snip/312b9f75-0e6a-4ace-8466-c48b2cd9f0f3)) - **Diverse Early Customers Validate Use Cases** - Early customers include AI video editor Mantra, AI search S-C CIRA, Cluey, and real estate startup Retz. - These customers show diverse use cases across video, search, and productivity apps. Transcript: Jaeden Schafer So a bunch of different companies are using him, and I think this is a fantastic tool with a ton of uses. But what I will say is this isn't the only company. There are a bunch of other competitors in the memory space. There is Leta, there's Memo. Actually, I think Shaw worked there for a small amount of time. They're both building memory layer agents. And Sousa Ventures has actually invested in memories.ai, also Samsung invested in them. And that can tap into thousands of hours of footage to get insights. So that one's more in specifically like the video industry that they're working on. But there's a whole bunch of these different tools in different verticals. And so when he was asked about this, he said, more and more AI companies will need a memory layer. Super memory solutions provide high performance while allowing you to surface relevant context quickly. So basically he says what sets them apart is their low latency. There's going to be other competitors. I mean, I think that makes a lot of sense. I love this. I love that. It's not just focusing on video. He's focusing on like the chat bots and the chat bot related tools, personally, something that I'm interested in with AI box. So I definitely am looking into this right now and yeah, really exciting and a huge congratulations to the whole team over at super memory. ([Time 0:08:32](https://share.snipd.com/snip/aa04b721-47a8-45bd-8d2a-08b66dd15f43)) - **Competitive Landscape And Differentiator** - The memory layer market has multiple competitors focused on different verticals. - Super Memory aims to differentiate with low latency and broad multimodal support. Transcript: Jaeden Schafer But what I will say is this isn't the only company. There are a bunch of other competitors in the memory space. There is Leta, there's Memo. Actually, I think Shaw worked there for a small amount of time. They're both building memory layer agents. And Sousa Ventures has actually invested in memories.ai, also Samsung invested in them. And that can tap into thousands of hours of footage to get insights. So that one's more in specifically like the video industry that they're working on. But there's a whole bunch of these different tools in different verticals. And so when he was asked about this, he said, more and more AI companies will need a memory layer. Super memory solutions provide high performance while allowing you to surface relevant context quickly. So basically he says what sets them apart is their low latency. There's going to be other competitors. ([Time 0:08:39](https://share.snipd.com/snip/baf6db03-50e5-4bb3-bd7f-dcdce30eede7))