CEO & Co-Founder, Adept Ai
Search every verified David Luan interview, podcast appearance, and on-the-record quote — each transcript cross-checked by AI and human review to confirm speaker identity. David Luan, CEO and co-founder of Adept AI, has been discussing the company's focus on building AI agents for enterprise knowledge workers. He has described Adept's approach as training a foundation model that can translate natural language into actions on a computer, rather than competing directly with general-purpose LLM providers. Luan has stated that the company is "very enterprise focused" and is working on larger engagements where Adept agents can accelerate knowledge work. He has also emphasized that Adept controls both its own foundation model for agents and the product enterprises use, describing this as a bet on vertical integration. In various interviews, Luan has shared his views on the broader AI landscape. He has argued that the business of training large base models is "quickly commoditizing" and that companies in that space will need to find alternate ways of making money beyond having better models. Luan has also commented on the importance of reliability for agent products, stating that if an agent makes operational errors a third of the time, people will stop using it. He has predicted that AGI is "really not super far away" but that it will not diffuse through society quickly due to other bottlenecks. Luan has also expressed concern about regulatory capture, saying that lawmakers' lack of understanding of the technology makes it easier for those with ulterior motives to shape policy.
“DeepSeek was incredibly good work — it’s part of this broader story arc of first making ML systems smarter and then making them more efficient. Just because you can make more intelligence happen at a lesser price doesn't mean you stop consuming more intelligence; if anything you consume even more of it.”
“People will train humongous teacher models on as much compute as they can get their hands on and then try to figure out internally how to render it down to something that runs really fast and is efficient for customers.”
“An LLM trained to do next-token prediction by definition is penalized for discovering new knowledge because new knowledge wasn't in the training set; what we needed was to combine LLMs with paradigms that can discover new knowledge like RL and search.”
“We've always seen that these models are better at determining whether they've done a good job than at generating the answer, and to some extent what we're doing with RL is exploiting that to force them to try over and over to satisfy their own sense of whether they did a good job.”
“LLMs out of the box are behavioral cloners — the moment they end up in a situation they've never seen before the generalization tends to be bad. Reliability remains an issue: if this thing deletes a third of your QuickBooks entries one in seven times you'll never use it again.”
“To turn a base multimodal model into a large-action model you have to solve two things: an engineering problem — expose in a model-legible way what it can do (APIs, UI affordances) — and a research problem — teach it to plan, reason, replan and follow user instructions.”
“I think there will be specialized models not for technical reasons but for policy reasons — some companies just don't want their data commingled, and AI systems powering different divisions (like sales & trading vs. investment banking) shouldn't share information even remotely through their weights.”
“I personally think AGI is really not super far away; it will be deeply transformational but it won't diffuse through society quickly because other bottlenecks (Amdahl's‑type limits) will slow useful adoption — we'll have a capability overhang.”
“Amazon is super serious about building generally intelligent systems — people at Amazon understand that computing itself is changing and a call to a large model or large agent will probably be the most important compute primitive in the future.”
“One thing Nvidia did that doesn't get talked about enough is bringing interconnect in‑house and choosing to orient the business around systems — that was a really good move.”
“I think the stuff is just incredibly powerful, but we're still underestimating how much there is left to run on this stuff — it's still so incredibly early. When we were training these models at Google it became clear you could take hand‑engineered ML models, replace them with a giant model, fine‑tune, and get somethin...”
“If you ask generative models to do something for you in the real world they kind of just pretend like they're doing something because they don't have a first‑class sense of actions or what affordances are on your computer. What I'm excited about is: how do we train a foundation model of all of the actions that people t...”
“I think the default path that we're on is increasing scale and increasing data and I think that will continue to lead to a lot of gains — but the question becomes how do we pull forward the future faster. On the data side I don't think most people realize we're much more constrained on data than we think.”
“It starts with having a pretty crisp definition of the end game. For us we want to build teammates and collaborators for people — software tools that help humans increase the level of abstraction at which they interact with their machine. It doesn't replace the musician, it gives musicians synthesizers.”
“The root thing founders should ask is what loop your company will run to compound — is it deep customer understanding or a data flywheel? I don't think we want to live in a world where companies are outsourced customer‑discovery engines and then Amazon‑basic versions of those things come out over time.”
David is an OG in AI who has been at the forefront of many of the major breakthroughs of the past decade. His resume: VP of ...
A re-broadcast of Greylock general partner Saam Motamedi's interview with Adept CEO and co-founder David Luan and Stanford ...
David Luan is the CEO and Co-Founder at Adept, a company building AI agents for knowledge workers. To date, David has raised over $400M for the company from Greylock, Andrej Karpathy, Scott Belsky, Nvidia, ServiceNow and WorkDay. Previously, he was VP of Engineering at OpenAI, overseeing research on language, supercomputing, RL, safety, and policy and where his teams shipped GPT, CLIP, and DALL-E. He led Google’s giant model efforts as a co-lead of Google Brain. ----------------------------------------------- Timestamps: (00:00) Intro (01:03) Lessons from Google Brain & Their Influence on Bu…
Guest episode of Product-Led AI, a new podcast series hosted by Greylock partner Seth Rosenberg featuring conversations with leading AI builders. Adept CEO and co-founder David Luan has been riding the LLM wave since the early days of the modern AI boom: an early OpenAI employee, he led the company’s engineering team before serving as the tech lead for Google Brain’s work on LLMs. He co-founded Adept in 2022 to develop multimodal agents built to work alongside humans in any profession. Believing computers should work more like humans and not the other way around, Adept has stood out for its…
Adept CEO and co-founder David Luan has been riding the LLM wave since the early days of the modern AI boom: an early ...
2024 is the year of multimodal models, which understand images as well as text—and the first year that AI agents are becoming mature enough to be useful. But are they enough to reach artificial general intelligence (AGI)? Join Bryan Catanzaro, VP of Applied Deep Learning Research at NVIDIA, and David Luan, CEO of Adept AI as they discuss the road to increasingly useful capabilities in the future of AI and the path to AGI. Panelists: ▫️ Bryan Catanzaro, Vice President of Applied Deep Learning Research, NVIDIA ▫️ David Luan, Founder and Chief Executive Officer, Adept AI and former head of en…
David Luan has been at the center of the modern AI revolution: he was the ~30th hire at OpenAI, he led Google's LLM efforts and ...
Delve into Adept Co-Founder and CEO David Luan's perspective on the transformative role of AI in shaping our digital and ...
Turns out, training AI is a lot like onboarding a new hire at work. 🤯 Want to teach AI to better support your team? Join us on Thursday, Feb 29 for the first-ever Masters of AI Day! On AI Day, you’ll learn how to build an AI-powered team through a hands-on workshop featuring prompt engineering, collaborative activities, fun experiments and more in our curated AI Activity Pack. Learn more and register at the link in our bio. Learn more and register here: https://waitwhat.mastersofscale.com/m... #MastersOfAI Follow the Masters of Scale podcast: https://listen.mastersofscale.com/follow Subscr…
Adept's David Luan wants to build an AI teammate for everyone. When he was only eight years old, he enrolled in computer science college courses - those early moments in that lecture hall charted his path. More than twenty years later, and he's stayed true to the course. He's now building useful AI systems at Adept that "enable humans and computers to work together creatively." Last year, on the heels of GC's lead investment in their $350M Series B, he sat down with GC's Deep Nishar to discuss his founder journey, where value will accrue in AI, the most interesting opportunities in the spa…
Sign in to search the full transcript archive, filter by topic, and access every quote from David Luan.
The summary and quote tags on this profile are produced with AI assistance from verified, first-person interview transcripts, then checked by our team to confirm the speaker's identity and the accuracy of every quote. See how we verify →