Lightcone Podcast: AI Startup Ideas
YC partners explore startup sectors being revolutionized by LLMs, the challenges and opportunities for new AI companies, and how the landscape is shifting.
The latest YCombinator Lightcone podcast episode, LLM-enabled Startup Ideas, discusses how large context LLMs are making previously unviable startup ideas feasible. The Lightcone series is now about 27 episodes deep, starting around mid-2024.
Hosts for this episode:
Garry Tan: Is the President and CEO of Y Combinator and also a Group Partner. Garry has a deep history with YC, having been a partner from 2011 to 2015 where he built key parts of the founder experience, and was also a YC founder himself with Posterous (YC S08), which was acquired by Twitter. He co-founded Initialized Capital, a successful venture fund, and is known for his extensive experience as an investor, his significant online presence with a large YouTube following where he discusses startups and tech, and his background as an early designer and engineering manager at Palantir.
Harj Taggar: Is a Managing Partner at Y Combinator with long and varied YC history, first joining as a partner in 2010 after his own startup Auctomatic (YC W07) was acquired. He later co-founded Triplebyte (YC S15), a technical hiring platform, before rejoining YC. He also co-founded Initialized Capital with Garry Tan and Alexis Ohanian (who is best known as the co-founder of Reddit and Serena William’s husband.
Diana Hu: A YC Group Partner, was the co-founder and CTO of Escher Reality (YC S17), an Augmented Reality backend company that was acquired by Niantic (makers of Pokémon Go), where she then became the Head of the AR platform. Diana has degrees in Electrical and Computer Engineering from Carnegie Mellon, focusing on computer vision and machine learning, and previously led data science at OnCue TV (sold to Verizon).
Jared Friedman: A YC Managing Director, co-founded Scribd (YC S06), a large digital library and document-sharing platform where he served as CTO. Jared has been a YC partner since 2015 and is also an active angel investor with a portfolio that includes Cruise, Instacart, Ironclad, and Rappi. He previously studied computer science at Harvard.
Chapters
00:41 What startup ideas could not work before AI?
06:06 Technical screening products
07:35 Truly personalized education tools
09:48 Do better products automatically get better distribution?
14:41 Moats
16:08 The need for platform neutrality
17:40 Big Tech and AI
23:24 AI horseless carriages
25:14 Gross margins
30:03 Full stack companies
32:30 ML ops
37:14 Updated startup advice for the AI age
Topic: The New Viability of Old Startup Ideas, Especially in Recruiting
Summary: The podcast kicks off by highlighting that current AI capabilities, like million-token context windows, are enabling a new wave of startup ideas, many of which are old concepts that can finally be executed effectively. Harj uses his experience with Triplebyte, a recruiting startup, as a prime example. Before LLMs, building an evaluation engine for engineers was a multi-year effort requiring the creation of large, labeled datasets. Now, AI can handle the evaluation piece from day one, allowing new companies like Merco to rapidly build and expand. This opens up possibilities for transforming multi-sided marketplaces into simpler, more efficient models.
Notable Quotes:
Garry: "Every other week, we're certainly realizing there's a new capability, a million token context window in Gemini 2.5 Pro. It's just really insane right now. And the thing to take away from that though, is that we have an incredible number of new startup ideas, some of which are actually very old, and they can only happen right now." [0:00:42-0:00:52]
Harj: "One thing I've been thinking a lot about recently is what are types of startup ideas that couldn't work before AI or didn't work particularly well that are now able to work really, really well." [0:01:06-0:01:18]
Harj: "For us to have gone from like engineers to analysts to all these other things would have taken years because again, we had to rebuild the label data set. Um, but with LLMs, you can just do that on, you know, day one effectively." [0:03:19-0:03:26]
Garry: "What are marketplaces that are three-sided or four-sided marketplaces that suddenly become, you know, two or three-sided or now there are two-sided marketplaces like Duolingo that are, you know, a little bit under fire because they're sort of starting to say, actually maybe we're just going to use AI for, uh, the person that you're going to talk to in another language." [0:03:46-0:04:08]
Topic: AI in Technical Screening and Personalized Education
Summary: The discussion extends to specific applications like technical screening products. Companies like Apriora are using AI agents to conduct technical interviews, a task previously time-consuming and disliked by engineers. This not only saves time but also expands the market by allowing for more sophisticated and nuanced evaluations. Similarly, AI is unlocking the potential for truly personalized education. While the internet promised personalized learning, LLMs are making the "personalized tutor in your pocket" a reality. Companies like Revision Dojo (exam prep) and Edexia (grading assistance for teachers) are cited as examples.
Notable Quotes:
Diana: "Even with this uh recruiting idea space, this is company called Apriora that Nico, the other GP here at YC, funded back in winter 24. And their whole premise is to build AI agents that run the screening for technical interviews, where a lot of engineers spend a lot of time just doing a bunch of interviews and the pass rate is so tiny..." [0:06:13-0:06:24]
Harj: "But Apriora's product now with LLMs, you can do more sophisticated evaluations to kind of get more nuanced levels of screening." [0:07:15-0:07:22]
Harj: "We've never had really truly personalized learning or a personalized tutor in your pocket idea, which is possible now for the first time." [0:08:15-0:08:26]
Jared: "There's like a lot of studies that show that like the biggest reason that teachers churn out of the workforce is that they hate grading assignments. It's just like no fun at all. And so Edexia like is an agent that's like very good at helping teachers to grade assignments." [0:09:19-0:09:28]
Topic: Distribution, Cost of Intelligence, and Business Models
Summary: While AI enables better products, the challenge of distribution remains. The cost of intelligence is decreasing, potentially leading to a return of freemium models in consumer AI. If AI can deliver value comparable to human services (e.g., a tutor), consumers may be willing to pay significantly more than for typical apps, creating viable business models even without massive user numbers.
Notable Quotes:
Harj: "Do better products automatically get more distribution or will these startups have to work equally as hard to get distribution to be big companies as before?" [0:10:12-0:10:12]
Garry: "The cost of intelligence is coming down quite significantly. So, you know, I know that we tease this sort of uh almost every other episode, but like consumer AI, it finally might be here soon." [0:10:52-0:10:55]
Harj: "And so if your app goes from being like a self-study course that doesn't get any completion to actually being on par with the best human math tutor for your 12 year old, parents will pay a lot more for that." [0:14:11-0:14:22]
Topic: Moats and Platform Neutrality in the Age of AI
Summary: The conversation touches on building defensible businesses ("moats") in AI, emphasizing brand and switching costs. There's a discussion about the role of big tech companies like OpenAI and Google. The need for "platform neutrality" is raised, drawing parallels to net neutrality and the historical need to prevent self-preferencing by platform owners (e.g., Microsoft with Internet Explorer). The example of Siri's perceived lack of advancement is used to argue for user choice in AI assistants on major platforms.
Notable Quotes:
Garry: "Like what you need is brand, you need switching costs." [0:14:48-0:14:48]
Garry: "I actually really need platform neutrality. So in the same way, you know, 20, 30 years ago, there were all these fights about net neutrality. This idea that there should be one internet that ISPs or big companies should not self-preference uh their own content or the content of their partners." [0:16:12-0:16:20]
Garry: "Why doesn't this exist for voice on uh phones? Like you should be able to pick, not you shouldn't be forced to use Google Assistant, you shouldn't be forced to use Siri, you should be allowed to pick." [0:17:21-0:17:35]
Topic: Big Tech's AI Efforts and Challenges
Summary: The speakers analyze the AI strategies and shortcomings of major tech companies. Despite Google's powerful Gemini models and TPU hardware, its consumer-facing AI products and integrations (like in Gmail) are seen as lagging or poorly implemented, possibly due to internal organizational issues ("shipping the org") or the innovator's dilemma. Meta's AI integrations in apps like WhatsApp are also critiqued for user experience.
Notable Quotes:
Harj: "I saw some numbers recently about how, um, Gemini Pro models, like just their usage, particularly from consumers, is just an insignificant fraction of chat GPTs." [0:17:40-0:17:54]
Jared: "Which is fascinating since Google already has all the users with their with their phones." [0:18:12-0:18:12]
Diana: "I think it's suffering a little bit from being too big of a company and essentially shipping the org." [0:19:33-0:19:38]
Harj: "It's like if Google replaced google.com with Gemini Pro, it would instantly presumably be like the number one chatbot LLM service in the world, but they would give up 80% of its revenue." [0:21:27-0:21:41]
Garry: "But, you know, you have this meta AI and you ask it, hey, who are my friends? I'm going to Barcelona next week. Who are my friends in Barcelona? And it's like, sorry, as an AI, I actually don't have access to that. It's like, what, you know, what is the point of this?" [0:23:10-0:23:20]
Topic: The Resurgence of "Full Stack" Companies and Gross Margins
Summary: The concept of "tech-enabled services" or "full-stack startups" from the 2010s is revisited. Many of these, like Atrium (law) and Triplebyte (recruiting), struggled due to poor gross margins and the difficulty of scaling operations-heavy businesses. However, AI now offers the potential to automate the "ops component," allowing new full-stack companies to achieve software-like margins. Lagora, an AI company for legal work, is presented as a modern example that could eventually become a full-stack law firm powered by AI agents.
Notable Quotes:
Jared: "What I've been excited about recently is like I think you can make a bull case that like now is the time to build these full stack companies because like, you know, like you were saying like the Triplebyte 2.0s won't have to hire this huge ops team and have bad gross margins. They'll just have agents that do all the work." [0:30:03-0:30:16]
Garry: "But that that wave of startups generally forgot that you need gross margin." [0:26:41-0:26:45]
Harj: "Eventually their agents are just going to do all of the legal work and they'll they'll be the biggest law firm on the planet. Um, and yeah, I think that's the kind of full stack startup that just wasn't possible pre LLM." [0:30:57-0:31:02]
Topic: The Evolution of ML Ops and the Importance of Persistence
Summary: The conversation shifts to ML operations (ML Ops). There was a period when YC was hesitant to fund ML Ops tools because the underlying ML wasn't mature enough to create a large customer base. Now, with the AI boom, companies that persisted in this space, like Replicate and Olama, are thriving because the demand for their tooling has exploded. This underscores the importance of founders following their curiosity and sometimes enduring periods of obscurity until the market catches up.
Notable Quotes:
Harj: "I remember a class of idea we weren't interested in funding was anything in the world of like ML machine learning operations or ML tools." [0:32:30-0:32:45]
Jared: "I think that was the core problem is that like these people were building ML tooling, but there was no one to sell it to because like the ML didn't actually work." [0:33:34-0:33:41]
Diana: "You actually have a team that stuck it out. I mean, part of the lesson is sometimes it will take a bit of time for technology to catch up. And this company called Replicate that you work with, stuck it out." [0:33:56-0:34:03]
Topic: Updated Startup Advice for the AI Age: Follow Curiosity
Summary: The traditional startup advice of "sell before you build" and extensive customer discovery, born from an era where good ideas were scarce, might be less relevant now. In the current AI landscape, where new technological capabilities are emerging rapidly, the speakers suggest that founders should focus on exploring interesting technology and following their curiosity. By living at the "edge of the future," they are more likely to stumble upon valuable startup ideas.
Notable Quotes:
Jared: "Back in the pre-AI era, it was really hard to come up with good new startup ideas because the the idea space had been picked over for like 20 years. And so a lot of the the startup advice that people would hear would be like, you you really need to like sell before you build." [0:37:32-0:37:48]
Jared: "But I would argue that in this new AI era, that the right mental model is closer to what Harj said, which is just like use interesting technology, follow your own curiosity, figure out what's possible and like if you're if you're doing that, if you're living at the edge of the future, like PG said and you're exploring the latest technology, like there's so many great startup ideas you're very likely to just bump into one." [0:38:06-0:38:29]
Garry: "You apply the right prompts and the right data set and a little bit of ingenuity, the right evals, a little bit of taste, and you can get like just magical output. And then that's still a secret." [0:38:29-0:38:47]
Concluding Thought:
Summary: The overarching message is that it's an exceptionally opportune time to build AI companies. Many ideas that were previously impossible or impractical are now within reach, and the best way to discover them is by actively building and exploring the rapidly advancing technology.
Notable Quote:
Harj: "My main takeaway from this has been there's never a better time to build. So many ideas are possible today that weren't even possible a year ago. Um, and the best way to find them is to just follow your own curiosity and keep building." [0:40:21-0:40:27]