Santi: Pieter Levels — PhotoAI — eighty-seven percent profit margins. One product. One person. Over a hundred K a month in MRR with thirteen grand in infra costs. That's the PLG micro-SaaS dream, right? Ship it, grow it, collect checks from a beach.
Kira: Brett Williams — DesignJoy — hundred and thirty K MRR. Also one person. But it's a service. Productized, strict boundaries, no calls, forty to fifty clients paying five grand a month.
Santi: Same ballpark revenue. Same solo founder. Completely different models.
Kira: And completely different lives.
Santi: What do you mean?
Kira: One of them can disappear for two weeks and revenue doesn't blink. The other one — if Brett stops working, the queue stops moving.
Santi: Right, but Brett's cash flow was immediate. Pieter spent months building audience before PhotoAI earned a dollar.
Kira: So which one should you pick?
Santi: That's the wrong question.
Kira: It's the question everyone's asking.
Santi: It's the question everyone's asking, and it's the reason they're stuck at twelve K MRR running three offers across two models and wondering why nothing's compounding.
Kira: If you're sitting between eight and twenty K MRR right now with two or three offers running — some services, maybe a tool you're building on the side — every week you stay unfocused is a week your margins erode, your channels compete with each other, and your capacity fragments across time zones you can't serve well. That's not hustle. That's drift. And drift is what kills nomad businesses between year one and year two.
Santi: Season two starts here. One path. Ninety days. Weekly gates that tell you whether it's working or not — before your next visa run forces the question for you.
Santi: So here's the pattern I keep seeing. Somebody in the community DMs me — they've got an automation agency doing, I don't know, ten K a month. Three or four clients. And on the side they're building a SaaS tool because they read a thread about how services don't scale. And maybe they've got a course outline sitting in Notion because someone told them to monetize their expertise.
Kira: I feel personally attacked.
Santi: You should. You were doing this in twenty twenty-three.
Kira: I was absolutely doing this in twenty twenty-three. Agency, course idea, a half-built Chrome extension. Three Stripe accounts.
Santi: And your MRR was what?
Kira: Eleven K. For like eight months straight. Couldn't break through.
Santi: Because every hour you spent on the Chrome extension was an hour you weren't closing the next agency client. And every hour you spent on the course outline was an hour you weren't building the systems that would let your agency run without you.
Kira: And the worst part — none of them were bad ideas. They were all reasonable. They just couldn't all get enough of me at the same time.
Santi: That's focus debt. And it's the number one thing I see holding people in that eight to twenty K range. Not bad products. Not bad markets. Just too many bets running simultaneously with one person's attention split across all of them.
Kira: Okay, so the prescription is pick one. But pick one what? Because the AI agency versus SaaS debate is everywhere right now, and most of the content out there is just vibes. "SaaS scales better." "Services generate cash faster." Nobody's showing the actual math.
Santi: So let's do the math. Three paths. Real numbers where we have them, honest caveats where we don't.
Santi: Path one — AI-augmented services. You're selling your expertise, but AI does sixty, seventy percent of the execution. Think automation agencies, AI content ops, voice agent setups. You're the strategist and the quality layer. The models do the heavy lifting.
Kira: And this is the fastest to revenue. I want to be clear about that. If you need to clear your Visa Run Revenue floor — your rent, your coworking, your flights, your buffer — services get you there in weeks, not months.
Santi: Agreed. The ARPU range we're seeing from founder posts — and I want to flag, this is mostly self-reported on Indie Hackers, so take it directionally — is three hundred to fifteen hundred a month per client for recurring AI service packages. Voice agents, automation builds, content pipelines.
Kira: So at a thousand a client, you need what — ten clients to hit ten K?
Santi: Ten clients. And that sounds manageable until you factor in the hours. Each client needs onboarding, QA, async communication, scope management. Even with AI doing the production work, you're still spending — conservatively — five to eight hours a week per client on the human side.
Kira: Five to eight? That's generous. When I was running my agency at that stage, it was closer to ten per client because half of them wanted synchronous calls.
Santi: Which is why the acquisition channel matters so much here. SmartClick Systems — they're a Zapier-certified automation agency — they get leads through the partner directory. Zapier Experts, Make partner listings. Those leads come pre-qualified. They already know what automation is. They're not asking you to explain what an API is.
Kira: Versus cold outbound where you're educating and selling at the same time.
Santi: Exactly. So the channel for services is partner directories, referrals, and outcome-driven case studies. Not SEO. Not content marketing. You don't have time for that when you're delivering for ten clients.
Kira: And the margin question — what's left after your LLM costs, your contractor costs, your tooling?
Santi: Depends on the stack. If you're running voice agents, you've got ASR, TTS, and LLM inference stacked on every call. Claude Sonnet is three dollars per million input tokens, fifteen per million output. That adds up when you're processing hundreds of calls a month. But for most service setups — content pipelines, automation builds — your AI costs are under two hundred a month per client. So at a thousand ARPU, you're looking at sixty to seventy percent gross margins before your own time.
Kira: Before your own time. Which is the whole problem.
Santi: Which is the whole problem. Your time is the constraint. And that's why this path works best for people who need cash now and are willing to trade hours for it — temporarily.
Kira: Path two is where you templatize the service. Same delivery, but with strict boundaries. One offer, fixed price, no calls, one request at a time. DesignJoy is the extreme version of this.
Santi: Brett Williams — solo operator, hundred and thirty K MRR, forty to fifty clients paying roughly five grand a month. No calls. Ever. One request at a time per client. Word of mouth and Twitter as the primary channels.
Kira: And he did that alone.
Santi: Alone. But — and this is the part people skip — the capacity ceiling is brutal. Forty to fifty clients at that service level means Brett is working constantly. We don't have verified hours-per-week data, but the structure implies a very full schedule.
Kira: Right, and the only way to grow past that ceiling is either raise prices — which he did, multiple times — or add people. Video Husky went the other direction. Similar model, productized video editing, but they scaled to a hundred and eighty-three K a month with fifty employees.
Santi: Fifty employees. So your margins crater. You go from DesignJoy's implied seventy-plus percent gross margin as a solo to — what, thirty, forty percent with a team of fifty?
Kira: If that. And now you're managing people across time zones, which is a whole different skill set. That's not a solo nomad operation anymore. That's a company.
Santi: So the productized path has a fork in it. Stay solo, cap your revenue, protect your margins and your freedom. Or hire, grow revenue, but accept that you're building an operations-heavy business.
Kira: And the channel here is different from services. DesignJoy grew on word of mouth and Brett's personal brand on Twitter. You need authority content — case studies, before-and-afters, a pricing page that does the selling for you.
Santi: Not partner directories. Not outbound. Authority and referrals.
Kira: Which takes time to build. So if you're choosing this path, your first thirty days are about shipping the offer page and getting five paying clients through your existing network. Not building an audience.
Santi: Path three. The one everyone romanticizes.
Kira: The one you romanticize.
Santi: I won't deny it. PLG micro-SaaS — product-led growth. You build a tool, people find it, they sign up, some percentage converts to paid. No sales calls. No client management. Revenue while you sleep.
Kira: Revenue while you sleep — after months of building with zero income.
Santi: That's the trade-off. PhotoAI hit fifty-four hundred MRR in week one — but Pieter Levels had a massive Twitter audience before he launched. That's not replicable for most people.
Kira: What about someone without the audience?
Santi: Danny Postma — HeadshotPro. He went the programmatic SEO route. Built hundreds of landing pages targeting long-tail keywords — "AI headshots for real estate agents in Denver," that kind of thing. Public claims ranged up to three hundred K a month in revenue, though we should note those numbers come from secondary sources and vary.
Kira: So the channel for PLG is either audience you already have or SEO you have to build.
Santi: Or Product Hunt plus affiliates for an initial spike. Podsqueeze — podcast content tool — hit sixty-five hundred MRR in two months using Product Hunt, SEO, and affiliates together. But that's an early benchmark, not a sustained growth rate.
Kira: And the margins?
Santi: This is where PLG shines. PhotoAI reported eighty-seven percent-plus profit margins. Thirteen K a month in infrastructure — Replicate API, VPS, other services — against six-figure monthly revenue. Once the product works, your marginal cost per user is tiny.
Kira: Okay but what about the risk nobody talks about?
Santi: Which one?
Kira: Arthur Mensch — CEO of Mistral — said publicly that up to fifty percent of traditional SaaS functionality could be replaced by AI agents. Bain published a whole analysis on whether agentic AI kills SaaS. If you're building a micro-SaaS tool today, you're building into a market where the floor might drop out.
Santi: That's real. And Bain's take is nuanced — they say the winners will be products that own deep integrations and proprietary data workflows. Generic tools get eaten. So if you pick this path, you need defensibility from day one. Not just a wrapper around an API.
Kira: Which means more build time. More technical depth. More months before revenue.
Santi: More months before revenue. But potentially much higher operating leverage once it works.
Kira: So we've got three paths. Each one has a different margin profile, a different time-to-revenue, a different channel fit. But there's a layer underneath all of this that most of the AI agency versus SaaS content completely ignores.
Santi: The nomad layer.
Kira: The nomad layer. Your Schengen clock is ticking — ninety days in any rolling hundred-and-eighty-day window if you're in the EU. Your wifi is unreliable. Your time zones shift every few weeks. These aren't inconveniences. They're operational constraints that change which path is viable for you.
Santi: I run everything through what I call the Lisbon Test. Can I operate this from a café in Lisbon with sketchy wifi and no synchronous calls required? Services — especially AI-augmented services with live client communication — often fail that test. Productized services with async delivery can pass it. PLG micro-SaaS passes it easily once it's built, but building it requires deep focus blocks that are hard to protect while traveling.
Kira: And then there's the cost side. Before you price anything, you need to know your Visa Run Revenue floor. That's your rent, your coworking, your flights and visa costs amortized monthly, your essential software, plus a buffer. For most people in our community, that number lands between three and five K a month.
Santi: And your pricing has to clear that floor after COGS. If you're running Claude Sonnet at three dollars per million input tokens and fifteen per million output, and your tool processes — say — fifty thousand tokens per user per month, your LLM cost per user is under a dollar. That's manageable. But if you're running image generation or voice synthesis, those costs multiply fast. Pinecone pods alone run about eighty dollars a month for a standard setup.
Kira: So you model the costs before you set the price. Not after.
Santi: Before. Always before. That's how you avoid the Bali Trap — building a business whose margins only work if you live somewhere cheap.
Kira: Okay, I want to push on something. Because the smart objection here is — why lock yourself into one path? Bain's whole argument is that the AI landscape is shifting fast. Services hedge against SaaS risk. SaaS hedges against service burnout. A portfolio approach gives you optionality.
Santi: And that's not wrong. Long-term, a portfolio can work. I run two businesses right now — a consultancy and a micro-SaaS. But I didn't start both at the same time.
Kira: When did you add the second one?
Santi: After the consultancy was clearing twenty K MRR consistently and I had systems running it without me for most of the week. Took fourteen months to get there.
Kira: So the portfolio came after the focus, not instead of it.
Santi: Exactly. The problem isn't diversification as a strategy. The problem is diversification as a starting position. When you're at twelve K MRR and splitting attention across three things, none of them get the concentrated effort they need to compound. You're not hedging risk — you're guaranteeing mediocrity across all three.
Kira: And the ninety-day window is short enough that you're not making a permanent bet. You're making a sprint bet. Pick one, run it hard, check the numbers every Monday. If it's not working by week eight, you have data to pivot — not a gut feeling.
Santi: So here's how you actually do this. You pick your path — services, productized, or PLG. You set your Visa Run Revenue floor. You budget your weekly hours — billable, builder, CEO time, and a buffer for the travel chaos that will absolutely happen.
Kira: And this is the important part — you set three weekly metrics. Leading indicators, not lagging. For services, that's qualified leads and proposals sent. For productized, it's net new subscribers and same-day request closes. For PLG, it's signups and activation rate.
Santi: Every Monday, you check three numbers. Did you hit them? Yes or no. Pass or fail. If you fail two weeks in a row, something's wrong with the system, not your effort. Adjust the system.
Kira: And the pledge — no new offers until the ninety days are up. No side projects. No "just exploring" a course idea. One path.
Santi: The hardest part for most of us.
Kira: The hardest part. But also the thing that actually moves the number. We built a template for this — the Leverage Map. It's got the capacity calculator, the channel-pick matrix for each path, the weekly KPI sheet with pass-fail gates, and a risk table for the nomad stuff — wifi loss, visa deadlines, API spend caps. It's in the show notes. Duplicate it, fill in your numbers, and you've got your next ninety days mapped before your next flight.
Santi: So — Pieter or Brett. Eighty-seven percent margins or hundred-and-thirty K MRR from pure service delivery. We started with that as a contradiction, but it's not really one. They both did the same thing. They picked one path and ran it until the numbers compounded. Pieter didn't hedge with a consulting offer on the side. Brett didn't pause DesignJoy to build a SaaS tool. They committed.
Kira: And that's the move. Not which path is objectively better — because there isn't one. It's which path fits your constraints right now. Your cash position, your technical depth, your Schengen days, your tolerance for client communication. Pick the one that clears your Visa Run Revenue floor fastest given who you actually are today. Not who you want to be in eighteen months.
Santi: Here's what I want you to do this week. Open the Leverage Map — it's on the Resources page. Duplicate it. Fill in your Visa Run Revenue floor. Pick one path. Set your three weekly metrics. And then post it. LinkedIn, X, wherever your people are. Tag the show. Say "I'm running this for ninety days." Make it public so you can't quietly abandon it in week three.
Kira: Week three is when the shiny new idea always shows up.
Santi: Week three is exactly when it shows up. That's why the pledge exists.
Kira: Season two. One path. Ninety days. We'll check in on how it's going.
Santi: See you Wednesday.