Y Combinator’s Dream Startup List Just Dropped. Here’s What Founders Need To Know (Clip)

So the the final topic we wanted to discuss today is that Y combinator the premier startup accelerator of course has released its regular request for startups. That's when the YC partners talk about the sort of startups that they would like to see formed and then applying to YC. They're saying we will preferentially admit startups in these areas into our accelerator because we think this is where the puck is headed.

Now there are 14 of them. I don't think we have time to go through all of them, but I thought we could pick a few. Um, and the first one, uh, is full stack AI companies.

I think Chris, this will resonate with you extremely strongly. Uh, so this is what it says on the web page. Suppose you believe that LLMs are now able to automate a lot of legal work.

There are two things you might want to do with that idea. You could build an AI agent and sell it to law firms. That's what most people do.

Or you could start your own law firm, staff it with AI agents, and compete with the existing law firms. That, my friends, is going full stack. I love that.

I want that. I love that. Yes.

Oh my god. This is the This is the way, my friends. This is the way.

Oh my god. Yes. These are the companies that are going to be the most successful.

They're going to create the most value. This is the definition of disruption, my friends. The definition of disruption.

uh you you know if you believe that you can materially impact the end user experience the business model the cost the waste the inefficiency of an industry stop trying to sell it to the industry or not I mean some people can do that whatever if that's your dream is to build enterprise software for lawyers please go do that um if your dream is to make legal the legal system the legal process just magically better you're not going to do that with existing law firms you're not you're going to do that by killing the legit the the old law firms with Acme Co. new law to uh powered by Silicon Valley tech and so yeah go go disrupt go forth and disrupt 100% and I would say that this is not a new model so if you look at lemonade they've been doing that for a very long while they took the insurance industry they took prediction models to uh to look at actuary and risk and they're basically they IPOed based on on that instead of selling software to insurance companies they became a very very popular insurance company. So, if you want to look at like a a working model for this, Lemonade is a great company.

I asked every startup that I talked to that tells me that they're using AI to make enterprise software in an industry. I told them, why are you selling to that business industry and not really disrupting that industry by going native? I think that that's very very important.

Amir, I've got two or three artifacts on my website you should share with these founders. Uh it explains the reasons that B2B is not less risky than B TOC uh and is the way to go for full DR disruption and um you know you mentioned lemonade it's I mean what about Uber? What about Netflix?

What about all the great unicorn? What about Google? Google tried to go to Yahoo and Yahoo told them to piss off.

Uh that's and luckily that thank thank god they did because they built a full stack disruptor search engine. uh you know, every major unicorn you can think of pretty much uh has gone and done this. Um I'm sure there are exceptions uh you know, whatever Shopify and uh and a few others, but uh yeah, go go sell your own Go disrupt the incumbents.

I love that this is the the first thing on the YC request for startups list is number one on the list. We will go to number two because I also think this is one of the more interesting ones which is more design founders. So the concept here is that designers actually have a lot of the most important skills it takes to be founders but perhaps you know because they've lacked certain technical skills that's limited their ability to to to actually take that leading role as founders.

Uh and again with AI AI of course is sitting underneath everything in this request for startups. There's an opportunity to empower more designers to become really excellent founders. And so the idea here is yes, what software can you build that will support design founders.

If I were to look at that, I would I would couple that with other with other statements like more GTM founders. I think GTM is becoming go to market. The ability to acquire users, the ability to get more people to see your user, the ability to grab attention is very important.

So I would say okay maybe less just two engineers in a garage but the designer is welcome as well but also the GTM also maybe the the revenue founder I I think there's a there's a tide that is shifting towards non-technical founders but I still think that like technicalheavy companies will prevail. So my statement here is that not just design a lot of non techchnical founders will be more valuable um but technical founders are going to be super valuable for the foreseeable future. But but here's the thing Amir I I I think design is special because design is close to product and product scales.

uh and I really believe that brilliant design can create real real differentiation and real differentiation is the same as innovation. Um you know big time differentiation is the same as innovation. And so I I do think there is a massive opportunity for incredibly beautifully designed products that are both aesthetically beautiful and functionally uh effective.

And I think that the the world is just far too utilitarian. And I think there's a huge opportunity there. I I agree with you that yes, other roles can become, you know, founders more easily now as well.

But, you know, I'm now I'm going to show my bias and my prejudice. I think a lot of GTM or growth or sales leaders don't know how to scale uh or certainly sales leaders. Uh and so they don't know how to build products.

They don't understand prioritization. They don't understand craft. They don't understand scale.

And so they if you get a sales guy running a startup with with some vibe coding, god I would be really afraid to see what they build and who they build it for. It would be a complete disaster. Um you know maybe there are some other roles you know product managers as as founders um uh and other skills that are closer towards the the product and scale end of the spectrum.

Uh and I agree with you more people more crafts people all over the shop being founders but I think design and designers are quite special. Yeah, look, we did an episode actually not very long ago, Chris, that we called, you should you should go listen to this one, folks. We call it nextgen founders, startups in the age of AI, right?

And what we talked about is the fact that what AI does is it makes execution cheaper, right? Like execution used to be the long pole. It used to be the thing sitting in the critical path that determined your cycle length.

And with AI, that is getting shorter, right? The execution is moving from idea to implementation becomes faster. And what that means is everything else on either side of the execution, the design and product on one end, go to market on the other end are elevated in importance.

That is where a lot of the competitive advantage will be. So you know I think you guys are arguing like it's either or. But I think yeah, good design coupled with good go to market and also just that that ability to that that nearly that raw intellectual horsepower of being able to iterate quickly and actually ideulate and move meaningfully with what you've learned which I think you know a a really good designer and you know we've had Yuki Yamashida on the show the the chief product officer of Figma listen to that episode as well.

Um but you know we talked about design and what real design is. It's not about making it look pretty. It's about making it work really well and that requires many iterations and it requires taste and a really deep understanding of how humans interact with things in in this case with software and so yeah like I I do think that real true designers in that sense as opposed to visual designers are going to get a lot of superpowers and if we can give them certain tools to enable them to flourish in this environment then yes I I think there's a lot of power in that.

If this is one among the first of the episodes you've heard of our show, guys, go back to the back catalog in the early days where you'll hear your nev poo poo the idea that taste and judgment is a real differentiator. Oh, off, Chris. But no, no, but what I'm saying is I have have scored a victory that I have convinced him all these many hours, all these many years that taste and judgment plays a critical role in You can tell yourself that if you like, my friend.

I can tell you that the best founder that I worked with, Stuart Butterfield, was all about design. Like his entire entrepreneurship point of view was make it amazing to an amazing experience. So I I kind of agree that designer could be a top essence to drive a company.

Yaniv is right. It's it's not or it's an end. Okay.

Now there's a bunch of I guess requests for startup here that I kind of bundle in that just like you know using the latest AI technology to make broken things less broken. So there's one around voice AI there's an AI personal assistant healthcare AI AI residential security and AI voice assistance for email AI for personal finance. all of these like they're different sectors but we're basically saying okay here's something in the world that is really important that is really valuable uh that doesn't work very well right now and we can use AI to fix it to me that's like yeah okay sure I don't have a huge amount of commentary to add but I don't know if either of you folks would like to comment on that on that bundle or that category of startup requests one one thing I will add is um I continue to believe the AI companies the the application layer AI companies that have the most defensibility over the long term are those that physically interact with the real world.

So there's a lab doing some some lab work. There's a a home you're renting out with Airbnb, an asset you're moving in the real world like an Uber. So you have to physically interact with the real world or that the the software is some kind of massively multiplayer game.

think of a social network or a or an actual video game um or a marketplace. Uh those are hard for an LLM to dream up at runtime and and propagate amongst a small group of people because the rules need to be set. They need to be common and they need to interact uh and or they need to interact with physical assets in the real world.

Uh and so when when I look at this list, you know, some of these don't meet that criteria, but some of them very much do. So these ones that you've listed off voice AI, you know, personal assistant, residential security maybe, um these feel like, you know, especially AI voice assistant for email. Um these feel relatively, you know, in the crosshairs for disruption by just a raw LLM.

Um you know, the the personal assistant, I mean, Google's going to do that. Siri's going to do that. Um voice assistant for email.

Gmail's going to do that. just uh uh and so you're going to want to hope that these um you know, Frontier models don't get really really good and displace you. Although, just a side note, Gemini is a piece of garbage and can't help me with my email at all right now.

But uh so yeah, I think the some of these are a little bit in the danger zone and don't qualify for for what I'm saying. And that's that's a problematic thing. I think I think the way to look at it is to remove the word AI from there and say, does this industry merit disruption in a fundamental way?

look at it from first principles without the AI glitter like do we have the right team is this a big enough TAM is this a big enough pain and if it is solve it with other AI or a for loop I don't care how you do it go and do and do the job so I think like all these things are a mixed bag of okay these are broken things but do they have the right uh time is the pain big enough is the willingness to pay there so I would I will do a re regix and remove all the AI I from all these uh call to yeah but air I I think the the point here is why cominator is saying hey our thesis like everybody else is that AI is a massive platform shift that therefore reshuffles the deck on winners and losers and it forces you to re-evaluate all problems through a new lens and new first principles and therefore the the reason to exist the reason for why now is AI and so the question is what inefficiency and waste is left over by the legacy players and by legacy I mean one generation ago you know SAS and consumer apps that were preai and how do you disrupt those you know immediately you know near-term legacy players using AI as the first principles differentiator but I I agree but hear me out they gave two examples one of them is voice assistance to to email and the other one is AI for health I think one is a crazy time for health is a crazy time and tons of display you Actually, you could dive into it and do AI just for surgeons and you would have a crazy time and voice assistant for email. I would ask myself like is that a real business or to your point it will be disrupted by Gmail and Microsoft Office because they're they're incumbent and they'll be stupid if they don't have a PM that's thinking about it right now. So even if you take the AI from that equation, you would you would need to look and see the at at the basic merits is this a good of enough of a business to to innovate in.

Yes. And I think the other point just that's complimentary to what you're saying there is don't get so distracted by the shiny AI to forget that you're basically building tools to solve real problems in the real world. That's some of it.

And you know you see marketing copy that's like this powered by AI and that powered by AI. And to your point it's like I don't give a what it's powered by. Does it solve my problem or doesn't it solve my problem?

And so yes, AI is an enabling technology, but if you and so you've got to be very much aware of it, but you don't want to be fixated on the tool. You want to be fixated on the problem and then be hyper aware of what new capabilities the tool gives you to solve the problem, but the problem is still at the center. My pick design does not have AI.

Sorry, my peak design doesn't have any AI and it helps me carry things from one place to another. delightful soft, but wouldn't it be better if it did have AI in there? No, but I don't want AI in my big design.

The advice of course is is good and we've given it on this show a million times, which is solve a problem, don't get, you know, follow the shiny objects. But this YC request for startups is the definition of shiny objects, right? It's saying AI has disrupted everything.

We want AI startups and we want AI for everything. Um, and and yeah, you have to evaluate is that a real opportunity? Is there real inefficiency and waste?

Again, I'm saying there probably is when you rethink that industry with AI first principles, you will find a way to reduce inefficiency and waste. The question I'm saying is not is there real inefficiency or waste that could be solved by AI. The answer is yes.

The question is, is there white space that is not going to quickly collapse by one of the hyperscalers uh or their winner takes all acquisition of one of these startups that and you're going to be left for in the dead left for dead. So I I think that's the the thing the problem I've been pointing out for, you know, many months now is the white space is unclear. Uh you need to move fast, get acquired or consolidated and exit quickly and and retire is probably the the advice.

Okay. So now there are two requests for startup uh that are around education, right? So one is called AI personal tutor for everyone and the other is simply the future of education which is really more about you know teachers and and how we can make teachers more educators and take away a lot of that admin burden.

Now education is one of those areas. I I I thought I'd single that one out because first of all AI is so clearly relevant and going to challenge and change a lot of things about teaching and also can be incredibly powerful. uh but also I think it faces systemic issues that are as strong or stronger even than healthcare.

And so I think it's a bit of a unique space to think about and and I don't know if you guys have thoughts on on how education should be changing in in the presence of this technological wave. I think there's a lot of disruption to be made in education. There's also a lot of money.

If you if you ever send your kids to college, you'll see how much money you you have to pay to to get proper education. And I think that needs to change and fundamentally and fast. But I I think that all all the a lot of things are stacked against you if you're entrepreneur in education.

Um it's hard to raise money when you're entrepreneur in education. Um getting your customers uh is hard. Your customers do not have a lot of money.

Um and you you fight for attention. I I've failed to see a lot of successful entrepreneurs loving and returning back to education uh startups. They do it once and they say I don't want to go into that industry again.

Um and I think that's a signal that shows that there's like fundamental problems there that I don't think startups are set up to to be successful there. I might be totally wrong. I looked at a few uh educational startups.

I'm just saying I'm running empirically speaking. They have hard time acquiring users. They have a hard time monetizing those users and they have a hard time raising more funds.

In the US, education and healthcare are fully broken. The inequality between the halves and have nots. The cost to get highquality education or healthcare uh is astronomical.

Um and the outcomes per dollar are really the ROI is not there. Uh and so if especially if you're in the US or you're targeting the US which many of the our listeners should be um there is a massive problem to solve. At the same time as you just alluded to Amir selling to healthare um incumbent healthcare providers and selling to schools sucks.

They're very conservative. They're very riskaverse. Baking into their workflow is a complete nightmare.

And so this leads me to the advice we were giving earlier. Um, consider what a full stack healthcare provider looks like or as much of the stack as you can afford to reasonably build. Imagine what a full stack disruptor school might look like.

Um, what does it mean to be um, a student in a post AI global world? That's the kind of startup that is most interesting to me. I've advised a number of startups that help colleges and schools and they're all stuck going through the administrators and the teachers and those people for the most part either lack imagination or lack any time to apply their imagination.

Um, and it's probably the latter. They're all well-meaning and they're all, you know, trying their best, but they don't have any time and any uh tolerance for risk. um and instead I've encouraged a number of them to just go sell directly to students and teachers um who can just pick it up and use it in their classroom the same way that Asana or Canva or others have been used and and I think that's um that's something to consider as well.

It's interesting like you know there's a few things here and my wife's a teacher full disclosure. Uh you know first of all teaching or education is a highly regulated industry in most countries. Uh we also have public education systems that can't simply be displaced by a commercial model.

Um there's also the fact that schools are about a lot more than academic education right there's sort of a socialization and civics aspect and and things like that. I think there are a lot of problems with schools. Not all of them can be solved with technology.

I think one bit of lowhanging fruit that is mentioned that that is maybe more solvable is one thing that my wife complains about more than anything else besides you is the admin burden. Right. We're we're in majority audio format now.

If you need to do your you give me the finger in in with voice. you Chris. There you go.

Is that better? You're welcome. That's my full accessibility friendly insult.

You're full stack. Full stack. Full stack.

Full stack. up. Okay.

Where was I going? So yes, the admin burden, right? So teachers want to teach and you know a lot of them not all of them a lot of them are very good teachers but generally with the passage of time they have more and more to do that is not teaching um and you know often when when funding gets cut to education it's the administrative roles that go first and I don't mean the like not the administrative roles in that that the bureaucracies that create more admin work but the roles that actually support the doing of administration they go away those support roles go away and so teachers are stuck doing more and more of this nonsense on their own, right?

You know, filling in reports and forms and, you know, even simple things like preparing class materials, distributing class materials. And so, I do think there is a non-trivial opportunity for startups to automate that stuff away. Um, it's not going to be your next trillion dollar company, but I think that there's more of an opportunity and maybe a bit less resistance because it's not going to fundamentally change how education is done.

uh but it could make it significantly more efficient uh and allow teachers to do what they are there to do what they love to do which is teach. By the way, same exact problem for health care providers. I think education and healthcare have a lot of similarities in terms of the types of systems that they are.

Yeah, that that's why I connected the two in my in my response is that they're both they have an extraordinary set of similarities and and they are each maybe education more so the silver bullet to improving humanity. Um if we can uplevel our kids' education, we will have more rational, more empathetic, more effective leaders and um and it'll make uh the world a much better place. And if they can stop being sick and stuck in hospitals and emergency clinics, um the quality of life will go way way up.

We've got two more I wanted to cover quickly. So the first is an internal agent builder, which which sounds boring, but I think it's actually quite a deep insight from from what I've seen, right? And this says, "All companies will soon have one thing in common.

Every employee will build their own agents to automate the repetitive part of their jobs. We'd like to fund founders working on the infrastructure they'll use to do that. " Why do I think this is interesting?

Even though it sounds a bit boring, I'm seeing every company in this mad scramble to say, "Oh, we've got to use AI to automate our internal processes and all of that. " And often, you know, these initiatives are led by non-technical functions, by operations. I've heard them often being led by HR even and what you end up with is systems that frankly as an engineer I I find scary right like there there is a big difference between saying hey let's sort of play with some AI agents and let us use AI agents to automate core parts of how we run our business even if it is internal uh and you know potentially reduce our staffing needs uh that's a production system right and I think people don't realize that and so I see a lot of icebergs that are going to be hit soon because yes, tools exist that allow you to create internal agents, but they they're not safe.

They don't have all of the guard rails and all of the safety nets that allow non-technical employees to automate their work. And so I think whoever cracks that uh is going to be addressing a huge market. That's part of the interesting part.

I think if you create a synthetic human, uh, Ryan Hoover wrote an article that I really like about like synthetic humans and if you create a synthetic human and create the guardrails there and create the safety there, then you actually have a very very valuable product. I think I think the problem the problem from an engineering perspective and from a product perspective is extremely complex. You have if you go into any enterprise you have end systems that you need to access.

And MCP is just like the tip of the iceberg for how do you access systems? You need to have write access and you need to have read access and you need to have permissions and you need to have authorization, authentication. You need to have workflows.

You need to deeply understand interpersonal connections and where the knowledge is based in the company and who to ask. So I think there's tons of complexity in building internal workflows and connecting to internal systems. But if you nail that and it's very hard to nail but if you nail that you have a very val very very valuable system.

It's it's comparable to Mulesoft uh Salesforce Muleoft that did that for big enterprises. I think there's a place where you could have meaningful core workflows uh run by I if you're talking about hardcore industry strength industrial strength rails that your massive insurance company is running on or your healthcare company then you need hardcore enterprisegrade agent infrastructure tooling developer interfaces blogging and all this sort of stuff but I think if you're talking about agent that are spun up on the fly to do work for you. The dream I have and and my intuition is that these are going to ultimately be no harder to spin up than a Google Doc or a Google Sheet or a Google Slide.

You're going to there's going to be Google Agent Studio or Office 365 Agent Studio, Copilot Studio. You're going to click new and you're going to give it a task and you're going to say go and it's going to say what tools should I use or hey who's the product manager of that and uh what permissions and guardrails should I stick to and people are going to be spinning up quote unquote agents with the same level of forethought and and complexity as they do to spin up a document. Uh now this may take you know days, weeks, months or years to get to that reality but I think that's where it's ultimately heading and I think it ultimately again gets baked into some of these core productivity tools.

Quite possibly right but I think well maybe it's baked into the core productivity tools. I think what you're describing is very difficult and that is why this request for startups resonates with me. I think it can be easy to think that we're there now.

But anybody who has come into a growing business that's got a whole bunch of in Zapia knows that, you know, there's a difference between having a bunch of like, you know, chewing gum and duct tape and having internal agentic systems that are reliable and scalable. And so I think that's really what we're talking about here. Final one, just a quick one cuz I I like the aspiration behind this.

And this is AI research labs. You know what this one says is OpenAI began life as YC research, our in-house research lab. I didn't realize that either.

Open AAI pioneered the concept of independent AI research labs. Open AAI is doing incredible work, but there are many unsolved problems in AI and still plenty of opportunity for new research labs. So what they're saying here is the foundational space is not tapped out.

that you know if OpenAI is likely to be the next trillion dollar company there is more yet to be done at that foundation level and that they would like to fund ambitious technical startups to do that. Um do we agree or not? I think it's a no-brainer.

I think like having research in places I think GPT is a great advancement towards GI but it's not general intelligence and there's multiple places where we still have big gaps in knowledge that require deep research things like uh robotics things like what is the next level of intelligence that is not uh based on transformers um there's tons of places in this industry that is that are still like totally unknown uh how do you get consistency in AI in a in like a predictable way? There's tons of problems that if you do deep research in them, you will uncover extreme value and that's that's a good investment to make. Yeah.

Um I think this is obviously true and it's obviously a good thing to to put a call out for. Um, I just want to put a a point of caution out there for founders because I've bumped into a number of founders recently in my advisory work where they're positioning themselves as doing deep research but wanting to stay arms length from commercialization. They're like, we just want to be a research lab and we have a 2, three, five, 10 year time horizon.

Um, and we're raising millions of dollars to kind of get halfway through step one. Um and we don't want to touch the productization of this thing. And my advice has been to caution that, you know, we talk about full stack disruption, right?

Um you know, my religion, one of my religions is full stack disruption. If you build something truly novel that breaks some aspect of the world, chances are you need to take it to market. And so OpenAI is not a pure research lab.

It's a it's a deep tech product company. So I just want to caution some people's desire, people who tend to come from academia or come from deep domain expertise to want to live in a research lab but don't want to touch the commercialization. I I don't know that there's a lot of appetite for that and don't confuse this call for startups to be that.

That's not my perception. It's more like open AI. Let me take the other side of that because I think OpenAI is a unique organization from a unique point in time and for a long time they were a research lab that's why they started out as a not for-p profofit and they kind of you know through some combination of luck and good timing stumbled into this incredible product opportunity.

I don't particularly believe that lightning is going to strike twice in that way. And what I'm actually thinking about as we talk about this is I've got a friend who's a biotech founder, right, in like uh early drug discovery. And the way the biotech, this is also a VC funded space, right?

But the way biotech works is that the startups will never have the resources to take a new drug to market because of clinical trials and all of that stuff. And so the model is you basically, you know, you have a number of barriers are de-risking same as as in tech, right? where you're like de-risking some of the key parts of that drug discovery and you might get halfway to your point of mirror to taking this thing to market and at that point you get yourself acquired by big farmer.

So the life cycle there, the exit there is always nearly always acquisition by big farmer. And I wonder if actually that is what we're going to be looking at where these AI research labs are going to do fundamental research on pushing the frontier of technology and you know they're going to have trade secrets. They're probably going to be patenting things which I know Chris we normally advise startups not to do.

And then when they've got some technology that is at a sufficiently advanced proof of concept, they'll get themselves bought by OpenAI or by Google or whatever for, you know, a few hundred million dollars, even a few billion if they're lucky. And that will be the life cycle. Yeah, I think aiming for acquisition is one thing.

Aiming to be sort of a pure play research lab for the rest of time feels like a bit of a a difficult thing to pull off. Um, not not impossible. There are, you know, people have pulled it off.

uh it's just truly novel things require novel go to market and novel business models and and require you to typically pursue that to the to to the logical conclusion. Okay. Well, I think it's a really interesting one.

Uh these these YC requests for startups sometimes, like you say, they're a little bit shiny object syndrome, but they definitely point to the direction things are going in. A lot of really smart YC partners reading the tea leaves and seeing what they are seeing in tech uh from a front row seat. So, really fun to discuss.