Meet the Slowest Startup Incubator in the World—Pumping Out Billion-dollar Companies

Sam and Dan, welcome—so glad you’re here. I’ve admired how you run Boltin and Watt down the street in Brooklyn, and I love that you call it the slowest incubator in the world because you actually do it well.

We were pumping each other up this morning like, should we even do this? For context, we see two kinds of companies worth building now: AI‑native that push the frontier, and AI‑durable that use AI without changing their core machine.

I’ve known you both since the old New York tech days, and you even helped reconnect me with my future cofounder at a party years back—our paths keep crossing in funny ways.

Give us the model—why slow, and how does it work?

We start a company every couple of years in a niche that blends software, services, and a real‑world component, grow it to roughly five to ten million in revenue, then recruit a CEO better than us for the next ten times while we stay on as true partners. Examples: Moxie, which helps nurses launch and run compliant med spas, and a contemporary funeral service that’s online-first with events hosted at existing venues—now leading the California market and expanding.

Where’s Moxie now?

Series C, tens of millions in revenue, 600 plus customers, about 200 people—and it launched right before ChatGPT, so it wasn’t conceived as an AI play.

Why do the glass‑eating zero‑to‑one yourself and then hand it off?

That’s where the fun and the value are—figuring out signal with scarce data, pushing the boulder until it moves, and building the early systems; it worked twice, so we’re sticking with it.

I like avoiding the years four through ten existential grind by focusing on the early physics of a business, or by the time that stage arrives, spreading that responsibility across multiple companies.

How do you divide the work?

We tag‑team CEO roles—one runs the current company while the other starts the next—stay fully in each other’s context, and push each other while leaving room for our own styles.

We’ve also added partners, including one focused on idea generation and validation so we roll off one business into a warm start on the next.

Our speed is gated by two things: one great idea per year, and top‑tier talent from early hires to CEOs.

How does your model evolve with AI?

We like niches that are operationally hard and less crowded, and we treat AI as an accelerant; med spas and funerals don’t flip because of AI, but AI can speed execution around the edges.

Dan, where’s your head on AI now?

We favor AI‑durable right now—no robotic injectors for a long while—so we aim to be excellent users of AI for marketing, communication, and operations while avoiding the overly crowded AI‑native battles.

Let the geniuses fight in AI‑native; we prefer places with fewer knife fights.

What real efficiency wins have you seen, and what hasn’t worked?

Company discovery has been transformed: AI crushes the initial desk research and our Notion agent—Matthew Bolton—preps interviews, tracks hypotheses, and updates evidence from call transcripts; synthetic ‘AI customer’ calls were uselessly sycophantic, so we still need real conversations.

In one P and C insurance deep dive, the agent reread our POV, hypotheses, and the latest ten calls, then surfaced quotes and graded each belief so we could stay fact‑based while keeping the necessary founder mania in check.

We also use it to sharpen counter‑arguments, which keeps us honest the same way we do for each other.

AI gives us facts and speed, but the sparks come from our earned pattern‑recognition—like what really makes a business‑in‑a‑box work from building Moxie and talking to dozens of teams.

Out‑of‑the‑box business‑in‑a‑box ideas from AI rarely meet our criteria until we guide it with those learned properties.

How are you bolting AI into existing businesses?

We don’t reward ‘using AI’; we reward the best outcome given AI exists, so we seed great internal examples and raise the floor—fast greenfield wins like AI‑drafted landing pages and throwaway coverage tools, while core engineering still requires deployment and maintenance discipline.

Our talent team even uses SAMGBT, trained on my writing, to do first‑pass candidate outreach in my voice, which quietly boosted response rates.

Have the newest models changed things dramatically yet?

They’re better, but not a night‑and‑day shift for mature systems; the biggest speedups still show up in fresh, self‑contained work.

We treat chat search as another channel, not a revolution—people aren’t arranging funerals in a chat box, and even where AI drives traffic, it slots in like any other acquisition path.

Can you preview the next company?

Not yet—we just picked the direction days ago and need to build before we talk.

We’re hunting intersections where AI meets a secular shift, rather than AI as the only story.

I usually ship ideas as blog posts first and only turn them into businesses if they show legs—come back when you launch so we can dig into how AI fits; next time, let’s do it in person in Brooklyn.

I love your playful experiments; our first question is always who pays, so we’re weird in different but complementary ways.

This whole chat felt very Boltin and Watt, in the best way.

Smash like and subscribe to AI and I—this show is a treasure chest of wild insights and laughs, with Dan steering the ship into the future. Buckle up, because I’m head‑over‑heels for this ride.