Social AI: The problem Continua is solving
The $8M funding round that's changing group chat
Hi! My name is David. I left Google a couple years ago to start Continua at the forefront of the AI revolution. Recent AI chatbots are incredible in their abilities and have changed our lives. But they are conspicuously absent in the places where they can help the most: where people are already socializing and working together, like group chats.
That's why I'm thrilled to announce our $8M in seed funding from GV with participation from BVP and an all-star cast of angel investors. This funding is enabling us to tackle the hard problems of Social AI and help people across a wide domain of use cases. We have our Social AI Continua chatbot in general availability at continua.ai for anyone to try. In our capped initial release, Continua has already processed over 7 million messages and shown strong user retention.
My journey to Social AI
There's been a throughline in my career: my fascination with how machine intelligence can use context to proactively help people across phones and wearable devices. What do I mean? If your trusted devices know what's currently most important to you, they can give you information or make relevant suggestions without you even needing to ask.
I helped start Google Goggles (the predecessor of Google Lens) and was on the founding team of Google Glass (smart glasses). These systems "understood" the outside world, at least as was best possible before the advent of transformer-based LLMs. With this understanding, they could in principle remember where you left your keys, remind you what you need to talk about with a coworker, or suggest that something in front of you would make the perfect gift for a friend.
Before leaving Google, I was a Distinguished Software Engineer leading projects under the umbrella of "personal AI", that is, an AI that understands and represents a user, including screen understanding and privacy-preserving machine learning technologies. It was a time fertile with ideas. And yet I felt there was something in our ideas that was missing – I didn't know what – but that the startup structure in this fast-changing world would be the best structure to find "the next thing", the best structure to capitalize on the extraordinary AI progress that was changing the world.
Also, I had an itch to scratch. I was a serial entrepreneur at Google: I came up with product ideas, prototyped them, evangelized them, built teams around them, productionized them, launched them, and repeated this process. But it was always in an environment with near unlimited resources. I wanted to see how fast I could run on my own. Big Tech has "startup-like" programs that promise the freedom to move fast, but nothing is as exhilarating as the challenge of an actual startup where every minute counts and you are part of a small, dedicated team working under enormous resource constraints1.
The missing piece: social context
And so I left Google and started Continua with the broad goal to create technology that preserves and amplifies the most precious resource that there is: human attention. I put together a dream team of engineers and thinkers. Over time, we figured out what was missing to build a truly useful personal AI: an agent that works in social settings. Continua handles the logistics and remembers the details so people can focus their attention on what actually matters – each other. It's the difference between spending an hour debating where to eat versus actually enjoying dinner together. Between forgetting you haven't seen a friend in months versus Continua gently reminding you both to grab those drinks. Our goal wasn't to build an AI companion, but rather a tool that facilitates and enriches human relationships.
It's the difference between spending an hour debating where to eat versus actually enjoying dinner together
If Goggles and Glass assess the context outside of you, communications and group interaction represent your internal interests. If the outside world can be important context for AI to help you, what goes on across your social network, your plans, dreams, works-in-progress are a hundred times more important.
Bringing the power of AI to group chat is really hard, and is critically needed. How many times have you heard someone say, "this is what ChatGPT told me"? Wouldn't it be better if an AI were a quiet helper in the scenarios you are already in?
If you think about human progress over millennia, a lot of it has been about social interaction. It's not about one genius here or one genius there. It's about the social interactions that we have: being inspired, being motivated by what others do, being creative, being collaborative. Our brains have evolved and grown to support understanding other individuals and larger groups of people, which has led to incredible progress. Now we have a new intelligence in LLMs. It's not a human intelligence; it's foreign, but it has abilities that can complement ours and can be trusted. Having this new intelligence as an integrated and cohesive part of human-to-human group conversation is a real "unlock", and is at the heart of what our company is doing in Social AI.
Add the Continua agent to your group chats to make planning, coordination, and information retrieval effortless. No more playing 20 questions with your partner to find the right restaurant or walking your dad through steps on how to restart the WiFi. Continua assists everybody directly in the group chat, so you don’t have to leave the conversation, look things up, and copy and paste information back and forth. By offering relevant and timely assistance, Continua clears the way to let you focus your attention where it matters most.
Continua transforms group dynamics
The difference Continua makes becomes clear in everyday situations. When I planned a Disneyland trip with my sister, Continua didn't just help with the logistics beforehand – it was there in real time, helping us decide where to eat when we were hungry and which ride to hit next based on my daughter's changing preferences. No more scrolling through dozens of messages trying to find that restaurant someone mentioned three days ago. Continua participates in your group chats as a trusted member that:
Captures what matters: That brilliant hike suggestion buried under 200 messages of memes? Continua remembers it. Every decision, every important link, every plan that actually matters – automatically captured and ready when you need it.
Ends decision paralysis: We've all been there – five different restaurant suggestions, three people saying "I don't care," and two hours later you're still debating. Continua cuts through by instantly suggesting the best nearby options with reviews, hours, and directions. One member can simply say "Poll the group: pizza or Chinese?" and Continua handles the rest.
Creates living documents: As your group discusses projects, Continua automatically generates and maintains Google Docs. My own Continua-created docs include everything from "Bathroom Renovation Plans" to "Summer Upstate Getaway" to "Pet Allergy Management." These aren't static files – they evolve as the conversation progresses.
Never forgets, never sleeps: Set reminders for the group ("check if those concert tickets go on sale"), ask for updates ("any new AI developments related to our project this week?"), or have Continua periodically refresh information. The AI doesn't get tired, doesn't forget, and is always ready to help when the group needs it.
These features become richer in combination. For instance, you can ask it to "update our trip planning doc every week with new restaurant openings in the area" – it will track your conversation, maintain the document, research new information, and proactively share updates.
And here's something else that’s cool: you can direct message (DM) Continua privately to ask about things from your group chats without bothering everyone. Can't remember what restaurant John suggested last week? What time did we agree to meet? Just ask Continua directly – it knows the context from your shared conversations but keeps the side conversation between you two.
Meeting you where you are
There's no need to download a separate app to use Continua. Already texting your friends? Just add a Continua phone number to the group chat. Are you on a Discord server? Try this.
Leveraging our team's long experience in the industry designing privacy policies and technologies, we built an agent that you can trust, including encrypting data at rest and in transit. I personally have Continua in all of my group conversations, where it is a trusted member of the conversation that distinguishes between distractions and directions and gets smarter with every exchange.
The technical challenge
Not only does building Social AI address a real user need, it's also a worthy research problem.
The software engineering world is building amazing products on top of LLMs, some as thin layers and some completely retraining the model and applying it to very specialized situations. LLMs are instruction-tuned to work in the context of a single user and single assistant, where the user says something and the agent responds: a simple call and response. This is not sufficient for Social AI.
It is a hard problem to have AI engage in group interactions in ways that not only don’t degrade the interactions but actually make them richer. Think of having a two-way conversation, and then a third person joins. The conversation gets richer; the third person adds a new dimension, at the cost of losing the 1:1 intimacy. Now imagine Continua joining that three-way conversation. Our standard is that it should enrich the conversation just as if another person joined, bringing a fresh perspective without being disruptive. That is our bar for Continua joining conversations, and we feel that we’re the only ones even close to achieving it.
We often get asked, "what LLM do you use?" Well, we don't use just one! In fact, we have a number of fine-tuned models and filed patents. We sort of had to break the LLM's brain to get the kind of social intelligence and etiquette that is needed for it to be helpful to groups of people. This includes proactivity, where Continua itself will initiate chatting when it thinks it would be most helpful. You can read more about our technology in these other posts:
We sort of had to break the LLM's brain to get the kind of social intelligence and etiquette that is needed
Social AI is an emerging research area in which there are more breakthroughs to be made. And if you're interested in helping us push the product and technology further, please see our open jobs.
With gratitude and momentum
We've brought the power of AI to group chat. You can stop playing telephone by copy-pasting what an LLM says. Like a strong team player or a great friend, Continua will anticipate needs, offer actionable and timely assistance, and seamlessly integrate into the natural banter. I couldn't be more excited about the future of Continua as we build the world's leading Social AI! Where do you think Social AI should go next?
– David, founder and CEO
P.S.: Interested in keeping up? Subscribe to this blog and join our socials (X / Twitter, Bluesky, Instagram)!
Special thanks to our investors! The round was led by GV with participation from Bessemer Venture Partners and notable angel investors including Jeff Dean, Adam Ghetti, Eoghan McCabe, Josh Reznick, Peter Mattis, Spencer Kimball, Daniel Loreto, Ben Sigelman, Dor Levi, Zach Lloyd, Eyal Manor, Lauryn Motamedi, Des Traynor, Dan Shipper, and more.
