1:08:23Greg Isenberg · Interview
9 AI startup idea niches to make $50K/mo (Data-Backed)
Greg Isenberg and Jonathan Courtney pressure-test nine startup categories live and land on one portable rule: date the product, marry the niche.
May 18thThe argument in one line.
Staying solo and AI-native until product-market fit is not a sacrifice — it is the only way to understand the cutting edge well enough to win in the current era of AI-powered company building.
Read if. Skip if.
READ IF YOU ARE…
- A solo founder or micro-team building an AI product who is tempted to hire before you have clear PMF.
- A builder wondering whether SF physical presence still matters in a remote-work world.
- Anyone trying to understand how agentic infrastructure actually works — browser automation, GPU procurement, sandboxed agent loops.
- A founder thinking about go-to-market and whether feature naming affects virality.
SKIP IF…
- You are looking for a technical deep-dive into Pulsia's architecture — this is a vlog, not an engineering walkthrough.
- You want a balanced critique of the one-person-company thesis — the video is sympathetic to Ben's worldview throughout.
The full version, fast.
Ben Barocca built Pulsia — an AI agent platform that autonomously builds and runs companies — to $9.5M ARR with no employees after a $30M seed round. The video argues that staying solo through PMF forces founders to deeply understand every edge of their product, and that a partnership stack (infrastructure companies instead of headcount) can scale a startup further and cheaper than traditional hiring. GPU access emerges as the real ceiling between $7M and $70M ARR, requiring investor connections and multi-year reserved commitments to unlock.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Create a free account →Where the time goes.
00:00 – 01:42
01 · Cold open + intro
Ben's 2AM frustration with a coding AI sparks the Pulsia idea. Will frames the $30M seed and the one-person billion-dollar company goal.
01:42 – 03:15
02 · Morning routine and run
Will meets Ben for a run in the Marina. Founder discipline, rituals, and mental state as infrastructure.
03:15 – 04:45
03 · Pulsia briefing — the day's agenda
Ben previews the French podcast, the upcoming Boost autonomous-run feature, and flags the $7M ARR milestone.
04:45 – 07:45
04 · SF serendipity — Google Ventures coffee
Chance encounter with a Google Ventures investor. Ben explains why SF physical proximity compounds fundraising in ways no other city replicates.
07:45 – 09:30
05 · Walking to Sapiem HQ — the partnership model
Ben describes partnering with Sapiem, Anchor Browser, Blackcell instead of hiring. Working with other people without hiring employees.
09:30 – 11:55
06 · Go-to-market strategy session
Ben and the Sapiem team map PR channels for the fundraising announcement; debate rebranding Boost to God Mode or YOLO Mode for virality.
11:55 – 14:25
07 · Anchor Browser demo — agentic browser automation
Sapiem's Anchor Browser product explained: how AI agents interact with the web at scale without triggering bot detection.
14:25 – 17:20
08 · GPU problem deep dive
How GPU procurement actually works — reserved capacity, open-source model compatibility issues, black-market allocation dynamics. The $7M to $70M infrastructure ceiling.
17:20 – 18:29
09 · Founder advice and close
Ben's framework: stay lean and AI-native until PMF; only then replace employees. Understand the cutting edge or someone else will.
Lines worth screenshotting.
- Building a company to $9.5M ARR with no employees is now possible because AI agents can replace entire functional teams for a fraction of the cost.
- Staying solo until product-market fit forces you to understand the cutting edge — outsourcing that knowledge through early hires is where most startups lose.
- Partnering with early-stage infrastructure companies instead of hiring specialists gives you on-demand capacity that costs nothing when idle.
- SF physical presence still creates compounding fundraising advantages that fully remote founders cannot replicate — repeated casual encounters turn into relationship capital.
- GPU access is a black market: reserved multi-year commitments compete against buyers across the globe, requiring investor connections to even get an allocation.
- Renaming a feature from a technical label to an emotion-first one is a go-to-market decision that determines whether the story travels through media channels.
- Making VCs pitch your AI agent instead of pitching to them filters for investors who genuinely believe in the thesis — and is itself a product demonstration.
- The internet was built for companies running one persistent web server; agentic infrastructure requires entirely different economics — spin up in milliseconds, pay nothing when idle.
- A solo founder at $7M ARR is not a ceiling — it is a proof-point that justifies the next ten million in infrastructure investment.
- Understanding where the infrastructure edge is often reveals a better version of the business you were trying to build in the first place.
Stay lean until you understand the edge.
WHAT TO LEARN
Hiring before product-market fit does not accelerate learning — it outsources it, and in the AI era, the founder who understands the cutting edge most deeply wins.
- Staying solo through PMF forces you to understand every failure mode of your product, which often reveals a better version of the business than what you originally planned.
- Partnering with early-stage infrastructure companies instead of hiring specialists gives you on-demand capacity that scales with revenue and costs nothing when idle.
- SF physical presence compounds fundraising: repeated casual encounters with investors convert to relationship capital exactly when you need it for a raise.
- GPU access is not a commodity — it requires multi-year reserved commitments and investor connections to secure allocation against global competition.
- Renaming a feature from a technical descriptor to an emotion-first label is a go-to-market decision, not a branding one: it determines whether the story is shareable.
- A company can reach $9.5M ARR with no employees if the founder treats AI tools as the team and infrastructure partners as the org chart.
Terms worth knowing.
- Agentic infrastructure
- The layer of services — sandboxed browsers, API rails, GPU compute, memory layers — that AI agents rely on to take actions autonomously at scale, separate from the agent logic itself.
- Reserved GPU capacity
- A multi-year commitment to a GPU provider who runs the hardware and lets you choose which open-source model runs on it — similar to a hosting contract but for AI compute.
- Browser automation
- A service that lets AI agents interact with any website as if they were a human user, bypassing bot detection — enabling non-technical users to automate web-based tasks through an agent.
- ARR (Annual Recurring Revenue)
- The annualized value of subscription or recurring revenue, used here to gauge how quickly Pulsia scaled relative to traditional software companies at the same stage.
- PMF (Product-Market Fit)
- The point at which a product satisfies strong market demand, signaled by rapid organic growth and clear retention — the threshold Ben uses to decide when it is finally safe to hire.
Things they pointed at.
Lines you could clip.
00:09
“I'm talking to a fucking computer, and it literally will not fix that bug. And I'm like, what am I doing with my life?”
06:46
“It's sort of like that serendipity you get in SF where you keep on running into people casually — and then when you wanna raise a round, it's so much easier.”
16:52
“Until you have product-market fit, you should force yourself to be alone or a micro team.”
17:09
“The only way to win in this age is to fully understand the cutting edge. And if you start hiring too fast, that's where you may lose.”
Word for word.
00:02At 2AM with GBTV, go, why can't you fix this bug? And I was like, I'm talking to a fucking computer, and it literally will not fix that bug.
00:11And I'm like, what am I doing with my life? I'm gonna build an AI that builds companies, and then I'm gonna build a thousand companies with all my ideas, and the dream got really big.
00:23Everyone is building an AI in some way, shape, or form, but there's only a handful of people that are truly building at the fringe of what's actually possible. And today, we're getting to meet one of them. It's a super ambitious project.
00:35And so obviously,
00:36investors love really ambitious project that can change the world. It's also very impactful one if you can help the 99% be more of a part of the economy.
00:45I'm going behind the scenes with Ben Barocca, the founder of Pulsia, a platform that uses AI agents to autonomously build and launch companies for anyone. The mission of Pulsia is really to give a tool that is able to build any business, but also any idea autonomously, setting up a bank account, talking to an accountant doing taxes, to talking to a factory to build something specific, and then plugging into all of these to hiring a human.
01:09Ben just closed a $30,000,000
01:10seed round by making VCs pitch his AI agent instead of him. And his goal is to build the first one person billion dollar company. But the question on everyone's mind is, can he actually pull it off?
01:23What you really want is just like Bolsia to rip for you to tell us, Paulsia, hey, like, here's my business. Just do it. I don't wanna babysit you.
01:29Your AI just run for twenty four hours straight. This should be such a perfect indication of the fact that you can do this too. Anyone can do this.
01:38AI has just leveled the playing field. Go nuts and build something that the world's gonna use.
01:46Alright. So today, we are catching up with a guy called Ben Serra.
01:51He's a previously exited founder. He's onto a new venture called Pulsia. It is a agent architecture that allows anyone to basically type a sentence in and build a business and have it run completely autonomously twenty four seven.
02:06I meant to go on a run-in like twenty minutes. He's like a really good runner apparently. So he said he was not that good, which is what every good runner says.
02:15Probably gonna smoke me, so we'll see what happens.
02:34Life of a founder, it's like, I feel like you can't eat with the discipline because it's such a drive, and it there's so many ups and downs. You have to find those rituals because, like, you you only do your best work when you feel your best.
02:47Right? Yeah. And and So that's why, like, it's, like, it's hard for me to live somewhere that feels right to have a nice setup, like, a nice life setup when I'm working.
02:58My posture I don't know. Just trying to create the best experience for my customers.
03:02Okay. I also feel good. Yeah.
03:05I'm also in a good mental state.
03:08It's a beautiful spot.
03:11Cheers. Cheers.
03:14Super hectic run. Super hardcore.
03:17What's time today? What's going on? So today is gonna be about we have a podcast, uh, that we're gonna do in an hour.
03:24Oh, actually, the French a French podcast going international. Yeah.
03:29We're gonna release this new feature called Boost. Um, it's code named Boost. Don't I know if you wanna call it Boost yet.
03:35Uh, but, essentially, it's like a way for Polsia to run autonomously for much longer because cost of intelligence is so high. Like, you know, Entropic OpenAI APIs for the top models are so high.
03:46You know, the initial product was just like, um, you know, you had to you had one task every night, and then, like, in the product, you could ask to do more tasks every day one by one, and you can commission Polsia to just, like, run continuously for hours and hours.
04:08Alright. So we are on our way into San Francisco City to go to the office where Ben works at.
04:15He just finished up his podcast. It is absolutely crazy that within, what was that, like a month and a bit of actually like having this product live, that is at a 7,000,000 annual recurring revenue.
04:29That is just insane. It is like unheard of heard of numbers here, but it like in AI like it makes it possible. And when you see what he's actually building and what he's built for someone, like I'm like, oh yeah, I would use that 100%.
04:43Like this is the future. It's just insane.
05:01Can I be in the
05:03Welcome to the vlog?
05:06How do you guys know each other?
05:08We are French.
05:09French, Texin. Texin. She's a investor at Google Ventures.
05:13And
05:15Yeah. Yeah. We met in LA, I think, two Two minutes at a party.
05:19Two weeks ago at a party at one of your investors party, Sufja Morozo. Yeah. And the rest is history.
05:27Very excited about what he's building. I think you're really building the future.
05:33Some people may not see it now, but it's very exciting. Have you tried the sugary croissant next door from the Magri?
05:41Oh my god. Do you want one? Let me get you one.
05:43Do you want one? Sure. Okay.
05:45Fuck it. You're the best one. I'm telling him that, like, a fast everyday.
05:50Hello. Bonjour. Hi.
05:52In my Australian accent. Sometimes
05:59the universe wants you to eat croissant and you eat the croissant.
06:04We've been like we've been like fighting it for like we've been fighting it twice on two separate occasions this morning and it's it's happened. Tell me who that was, what just happened, like crazy proximity. Right?
06:14I know. So it's actually that's the
06:16that's an investor from Google Ventures that I met a while back in LA actually at like a party that like one of my investor threw back then. And yeah, when I moved to SF, she was one of the people that like I knew and then we hanged out and she actually she had the studio in Marina that she rented from me for a month.
06:33So I was in Marina for month. That's why I wanted to take you here. And it's such a cool, peaceful place to get coffee, to get the fourth coffee of the day.
06:40And, yeah, it's so random that I rent to her, but it's sort of like that serendipity you get in SF where you keep on running into people casually and then it's like, well, yeah, when you wanna raise a round, then it's so much easier when you're right there and you can just chat about it over coffee or over croissant.
06:57Right? Literally. Literally.
06:59And it's like if I was in Paris or in Houma or wherever, it's like that doesn't happen. Yeah. So you you can still build.
07:05You can still build great things, but you don't run into those people that like lift you up, that like give you opportunities to like go bolder, go faster, you know? Yeah. That's what you get at SF.
07:15That's what's a magical place. We're getting to Sapiens HQ, which is one of the infrastructure for agent sort of companies I work with to help me scale Postiere without having to, uh, rebuild the whole Internet myself.
07:30So, yeah, the idea is like partnering with the best SF AI agent startups that are all getting started, and are all building the the infrastructure for this future that's coming that's now and coming soon.
07:42Yeah. And yeah, it's like working with other people without hiring employees.
08:04The thing to think about it what would be interesting is sort of like having all these different channels. And so for each one, there's obvious big ones we could target. Yeah.
08:13And it's it's podcasters. It's like, you know, if it's a late night show, I mean, that's more difficult to do. But, like, at which diff at which stage of the company or which feature we think they could be interested.
08:26Because of course, we can wait until they invite us, but we can also provoke it with our network of people. But we can maybe today is not good.
08:33But if we announce the fundraising, maybe
08:36trying our luck to say, who knows? I think the fundraising, we should brute force it on each of those channels because the story is too beautiful to not being on all those things, those channels.
08:49The story is basically sort of founder company raising a shit ton of money
08:54and actually the company growing fast. And I think this should be one of the priorities from us, but also the PR agency Of course. To really be sure that we're going to be pushed on all those different channels.
09:05Yeah. Was thinking, for example, the solo founder journey. And that's something interesting.
09:10Those are obvious moments. Then there's features that could sound boring, but actually, we could turn it into an interesting story. Yeah.
09:18Right? So we talked about boost like the okay. Yeah.
09:20It's just a button to get Portia to be autonomous. God mode. But actually, no.
09:24Let's turn into let's call it god mode or yolo mode or, like Yeah. You know, like, something that's more interesting than boot.
09:33It's pretty cool to hear how Ben is actually thinking about go to market virality and customer acquisition, and that is one of the most critical pieces in getting a company like this to a venture scale outcome. But building a product that can actually deliver on that promise is a whole other challenge, and that comes down to the infrastructure partners making it all work behind the scenes.
09:54And that was exactly where we were headed next. Polsia, obviously, is building an AI that can build and run businesses for for the 99%,
10:04which means we get a lot of nontechnical people that, like, go to Porsche and ask them for their dream company or their dream projects. And
10:13sometimes Porsche needs to use a browser to act and to do things. For those who don't understand, would you mind introducing exactly what the product is? Yeah.
10:22So Encore Browser is basically an gigantic browser automation workflow that allow you to basically automate anything you want on the web without the hassle of, uh, are you human or are you bot and do it as fast as you can.
10:39So that's basically what we do, and that way we allow agents to interact
10:45with the web in a seamless way. Polsia is super pumped to be partnering with, you know, with Sapient, with Anchor Browser, with Blackcell, and a and a few other companies we're talking to to get everyone to benefit from each other.
11:00Right? Because everyone is is is building a part of the stack, going very deep in a part of the stack. And, you know, like, Polsia is, like, building more, like, the the the end product, like, the autonomous loops, the orchestrations, the, like, memory layers, and, obviously, like, the consumer, you know, the consumer growth, right, which is a big part of, like, what Porsche does and educates people.
11:22It's so funny. Like, you know, you're a solo founder, but it still takes a village. Right?
11:28But, you know, you can be a CEO and raise all the capital and hire an army with, like, the Lord of the Rings.
11:36You know? It's like, okay. I'm proto.
11:40Like, guys The fellowship have the ring, but, like, I need, like, you guys to come with me because, like, we're going to Mordor.
11:55So right now, the infrastructure is quite a it's quite a challenge to build, but it's Yeah. Amazing to see that much apps being deployed in a few sec in a few milliseconds or even nanoseconds.
12:07It's insane. I mean, really cool. Like, I love that, you know, we started working on just this inbox, which, by the way, is a whole thing with, like, all the security challenges of those agents Yeah.
12:16Going wild and like trying to like break everything if they can. And so being able to contain them into the sandbox where it's like, you know, you stay here, you stay Yeah. The corridor.
12:26And also, you know, end of the day, it's about reducing cost for our policy, which means we can reduce cost for the customer and get more customers.
12:34Yeah. It's the new infrastructure, like, only pay on demand. Like, when you're using stuff, you pay.
12:39But otherwise,
12:40it's free. Exactly. So And it's like I feel like the Internet was built for you.
12:44I mean, it was built for humans and for Definitely. And all those like web servers and boxes. It was built for a company to build one web server, and so there's like minimum commitment and cost.
12:55And when it's just an agent who's spinning up, you know, we're gonna launch a feature soon where it's like a user can launch like 10 businesses in parallel and just like AB test them. Yeah. And it's like, well, I cannot pay like $10 a month for 10 personal accounts.
13:09That makes no sense. I mean, no. Because actually, these are temporary businesses that get built in like in minutes.
13:15Yeah. So super excited for you guys to to sort of like figure out. At some point, we'll reach the the the infrastructure problem.
13:23Yeah. Probably where we need to have more servers. Yeah.
13:26But that's our More data centers. Yeah. More data centers.
13:29Centers so that we can support all CS growth. Exactly. We need to find VCs or, I don't know, someone that is excited about this about base scale.
13:40I we met a a VC earlier
13:42today. She's she's she's excited.
13:46She's gonna finance it. Yeah. And then there's the whole GPU problem.
13:49Like, that's something we wanna tackle afterwards. Like, we started with a sandbox, but we we want to optimize the whole infrastructure Yeah.
13:56Network wise, storage wise, GPU wise also. Yeah. And so that's a whole ecosystem that we want to provide.
14:03Yeah. I love it. And at some point, we'll tackle the GPU problem with our own data centers.
14:09All the network is inside these data centers, so you have the most optimal sandboxing and agentic platform Yeah.
14:18To work with.
14:26Sapiom is like the infrastructure that makes Porcelain possible. Porcelain was imagined and vibe coded by me.
14:35But very quickly as the system scaled, it's like, okay.
14:39There's a whole infrastructure of, like, you know, how do I how do I make sure that, like, when Polsia agents run, uh, they're secure, it's scalable, meaning if we go from, like, 1,000 customers to 10,000 to a 100,000, it scales.
14:54How do we make sure that, like, agents have all the best APIs and at the better at the best cost and that they are allowed to use those scales? So that was sort of, like, the the in the initial partnership.
15:06I was like, well, you guys build the payment rails and the API rails, and I'll bring the customers. And it's like we win together. And and now we are, like, buying GPUs.
15:18And now we're like, actually, okay. So how do we go from from, like, 7,000,000
15:23run rate to 70,000,000 run rate? How is it going on the GPU side? Yeah.
15:27It's still a mess? Yeah. It's hard to get GPUs.
15:29And it's hard to make those GPUs work at, like so we had a good concession today with
15:36What are they saying? Are they saying, like I feel like it's literally their job to run models.
15:41I told them. And so it's like told them actually. Do you mean, like,
15:44the top open source model is has an issue running on your fucking GPUs? Like, it doesn't make any sense to me. Just so I can understand, like, for the audience who doesn't understand how this all works, so you're buying GPUs
15:57from companies. How does that even work? Are you physically building
16:03infrastructure out west to, like, hold the GPUs? No. So there is companies who are actually running those GPUs, and they cannot resell the GPUs with the model.
16:13So it's similar to Entropic. You know, when you go to Entropic, your API is like technically, Entropic is running GPUs to run their model, but there there are companies that actually kind of reserve and buy kind of GPUs that they will run for you, and you decide the model that you want to run on those GPUs. So you can decide which open source model you want to run.
16:30So they kind of run those GPUs, and then you book those GPUs for a year, two years, three years. Uh, and then it's your kind of rack on our servers of GPUs, and then you can just turn all your traffic.
16:42And that's at no cost. You just pay the GPUs. And then it's up to you to optimize it and to, uh, leverage all those GPUs.
16:49But, uh, today GPUs, it's like a black market. Uh, it's almost impossible to get, uh, every one of those big companies, and they just got sold right away.
17:00Uh, so we're always competing against people in Asia, people everywhere to get some allocation of GPU, which is quite quite interesting.
17:09So it's all about connection and, you know, putting investors to figure out how we can connect with the right founders and making the right connections to get some clarity.
17:19To get a computer, that's crazy. Yeah. Yeah.
17:22Is there any sort of advice you'd give a founder who's looking to do something,
17:30I guess, a bit similar like this? Like, what would you say to those guys out there? I mean, what I think is that until you have product market fit, you should force yourself to be alone or a micro team.
17:40Like, you and one other person and that's it. Anyone on the team and if it's just you, it's just you, should use AI all day long like Flock Code, Codex, Polsia, whatever you want. And once you have Falcon Market Fit, try to get to replace employees.
17:55And what I'm trying to say is that like it will force you to fully understand where's the edge. And because once you understand what the edge, it might give you ideas to actually the business you were trying to start in the first place. And it's the only way to being able to win in this age is to fully understand the cutting edge.
18:11And if you start hiring too fast and then you rely on someone else's knowledge to build, that's where you may lose. And so stay lean, stay solo, and go get them.
The Hook
The bait, then the rug-pull.
At 2AM, staring at a bug a chatbot refused to fix, Ben Barocca had the only logical response: build an AI that builds entire companies. That moment of frustration is the origin story of Pulsia — and Will Phillips spent 24 hours inside it.
Named ideas worth stealing.
17:00model
Solo-to-PMF Framework
- Stay solo or micro-team until PMF
- Use AI tools all day — Cursor, Codex, Pulsia
- At PMF, replace employee roles with AI before hiring humans
- Use deep cutting-edge knowledge to find the real business worth building
Ben's framework for building in the AI era: stay lean and AI-native through PMF, only scale headcount once you fully understand where the edge is.
Steal forAny founder pitch or content about solo building
07:45model
Infrastructure Partnership Stack
- Sapiem — GPU and API rails
- Anchor Browser — agentic web automation
- Blackcell — additional infrastructure
- You bring customers; they build the rails; everyone wins together
Instead of hiring specialists, Pulsia partners with early-stage infrastructure companies and contributes customer flow in exchange for capacity and favorable terms.
Steal forAny founder building on top of emerging AI infrastructure
How they asked for the click.
VERBAL ASK
18:29link
“Check out Polsia (description link)”
Soft — link buried in description, no in-video verbal CTA. The close is motivational advice, not a product pitch.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
OTHER LINKSAlso linked in the description.
Visual structure at a glance.
