There’s never been a better time to start an AI company. Not just because there are new ideas, but because the tech finally makes old ones actually work.
On the Lightcone, Garry, Harj, Diana and Jared talk through the kinds of startups that are suddenly viable thanks to LLMs—from full-stack law firms to personalized tutors to recruiting platforms that can finally scale. They share the patterns they’re seeing, the ideas they’re excited about, and what it means to live at the edge of the future, where breakthroughs often look like second chances.
If you’ve been waiting for the right moment to build, this is it.
Apply to Y Combinator: https://ycombinator.com/apply
Work at a startup: https://workatastartup.com
Chapters (Powered by https://chapterme.co/) –
0:00 Intro
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
40:19 Outro
Chapters (Powered by ChapterMe) –
00:00 – Intro
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
40:19 – Outro
wild how no one here has mentioned the book 22 dark truths
The goal of AI is to eliminate startups.
Great content as always.
The idea of selling before building the app is outdated makes me a little bit confused. I know that coming up with new ideas is much easier than before. However, we still need to know the customer sector, their problems and the UVP you're gotta provide, right?
After all, most companies are just designing better/cheaper/faster solutions for existing problems.
Does Jared mean that we can simply ignore the development overhead because it's very low at the AI era?
3:32 then people evaluated would create code with AI
We are building an Ai CMS that is spinned up in micro machines with pg+vector+celery+redis where agents can do things whenever something happens similar to wordpress hooks but actions trigger agents and they do what it makes more sense to do with the information available… and instead of an agency managing your site, an embeded agent takes decisions on the fly and can hot load pyhton, manage the db, and build whatever you need on the fly! We have a super early mvp, and we can spin up this micro agents that have a cms bundled together in a few seconds. Hope you take a look, we dumbed down our answers a lot for the YC application! We applied as Flowpress but changed our name to Veloz recently ❤🎉
Students cheats on Assignments with AI, Teachers grade Assignments with AI 😂
AI grading AI
“Hugging Face” is such a weird name for a company.
This immediately goes in gemini to summarize. And then I let it generate a critical response.
Interesting discussion, but it feels like we're watching a familiar cycle on hyperspeed. Many of these "new" AI opportunities sound like repackaged tech narratives from previous eras. The real question is whether this is creating genuinely new business models with proven revenue, or just accelerating the regurgitation of old ideas.
Tools like AI can summarize and spread these concepts faster than ever, but this can create an echo chamber. We risk mistaking the rapid exchange of buzzwords for tangible innovation. Is this true progress, or are we just getting more efficient at discussing existing processes.