He Built a $160,000 Product in 10 Minutes and Kept Selling

I dropped something into a coaching call the other day that I almost didn't notice myself saying.

We were talking about this guy I was coaching - he's building an AI automation product for law firms, trying to get his first pilot client, working out whether to hire a dev team or just try to code the MVP himself. Standard early-stage stuff. And I was making the case for him to sit down with Claude and just build the thing himself, rather than pay a dev shop $2,500 for a flow he could probably spin up in an afternoon.

To make the point land, I mentioned - almost in passing - that the day before, I had rebuilt LeadShark.

If you don't know: LeadShark was a lead database I built and sold. The original build cost me $160,000. Engineers, investors, the whole infrastructure. It had filters, exports, download functionality, the works.

I rebuilt the same thing in about 10 minutes using Claude. Same core functionality - filter leads, view results, download whatever you want. Fully integrated into ScraperCity. Done.

I said it. I moved on. We kept talking about the guy's law firm automation product.

But I've been sitting with that moment since, because I don't think most people understand what it actually means. Not just for me. For anyone building a product business right now.

This Isn't a Story About AI Being Impressive

There are ten thousand posts about how AI is impressive. I'm not writing another one of those.

What I'm actually saying is more specific and, I think, more important: the people who are dangerous right now are the ones who have stopped being impressed by AI and started treating it as infrastructure.

When rebuilding $160,000 of work in a morning is boring enough to mention as an aside - when it's a rhetorical point in a coaching call, not the headline of a press release - something has fundamentally shifted in the economics of building a product company.

The guy I was coaching was stuck. He had a dev team lined up to build a two-phase automation workflow for estate planning law firms, and they were going to charge him $2,500 for phase one, $2,500 for phase two, and then $750 a month to maintain it. He was trying to figure out how to get a pilot client before he spent that money. Reasonable concern. But he was framing it as an either/or: either get the client first, or pay the dev team first.

I told him there's a third option: sit down with Claude on Opus 4.5 - the latest model built specifically for heavy-duty coding and agentic workflows - and just build it yourself. Four or five hours of focused work. The model asks you for your API credentials, your hosting preferences, your integrations. It walks you through every step. You don't need to know Python. You don't need a computer science degree. You need to be willing to learn, and willing to treat it like a video game you're determined to beat.

If he can do that, his cost goes from $5,000 in dev fees plus $750 a month, down to whatever the API calls cost - which, for an automation workflow, is going to be tens of dollars, not thousands. His margin doesn't go from 50% to 70%. It goes from 50% to 99%.

That's not a marginal improvement. That's a different business.

The 60,000 Lesson I Already Paid For

I've written before about building cold email infrastructure from scratch - raising hundreds of thousands of dollars, hiring engineers, managing investors. And from a marketing standpoint, it was a perfect idea. The cold email king building a cold email tool? No-brainer positioning. Huge demand.

But software stability has nothing to do with sales ability. And when you're operating in the gray areas that Google and Microsoft actively hunt - high-volume sending, deliverability edge cases - the software breaks constantly. The companies that built those tools did so over years, with dedicated engineering teams, constant iteration. You can't just build it in two months and expect it to hold.

LeadShark was a different product - a lead database, not a sending infrastructure - but the lesson is the same: I spent $160,000 learning exactly what the build required, what the architecture looked like, what filters and features and data structures made it work. That knowledge didn't disappear. It just got much cheaper to apply.

When I sat down with Claude and rebuilt it in 10 minutes, I wasn't starting from zero. I was applying a $160,000 education with a $20/month tool. The AI compressed the execution. The judgment was still mine.

That's the part people miss when they think "AI is going to replace founders." It's not replacing the knowledge of what to build, why to build it, or who to sell it to. It's just eliminating the gap between knowing and shipping.

The Dev Team Problem (And Why I Told Him to Ditch It)

Here's the actual sequence of events in that coaching call, because I think it's instructive.

The guy - let's call him a Galadon Gold member who came back after a break - had been building an AI automation product for estate planning law firms. The pitch: automate the repetitive, time-consuming workflows that HIPAA-compliant law firms deal with every day. Document intake, client data processing, the forms lawyers fill out over and over. The product integrates with Clio, which is the dominant CRM in the legal space, and the HIPAA compliance angle is the hook because lawyers are rightly paranoid about AI touching client data.

Good idea. Clear niche. Real pain point. He'd already talked to a few warm contacts in the legal world and had one paralegal who was basically begging him to build it - she just didn't control the budget at her firm.

The problem: he had a dev team lined up to build the flows, and he was treating them as the only path to a proof of concept. But when I asked him to walk me through the actual hard costs - if he ate everything, what does this cost? - it came out to roughly $2,500 for the initial build, plus about $750-800 a month in ongoing software and maintenance costs.

And I just asked the obvious question: why can't you build this yourself?

Not rhetorically. Genuinely. Because if you can sit down with Claude for even a few hours and vibe-code a working version of this flow, you're not eating a $2,500 cost - you're eating zero. And your monthly overhead isn't $800. It's the cost of API calls, which caps out well below that at any reasonable volume.

He hadn't thought about it that way. He'd been thinking about AI as a research tool, a writing tool, something that helps you go faster at tasks you already do. He hadn't fully made the mental shift to: this thing can build the product for me.

The Real Unlock: Build It, Get the Testimonial, Then Scale

Once he made that shift, the whole strategy clarified.

Step one: build the proof of concept himself using Claude, specifically on Opus 4.5. Not a polished product. A working demo. Something that automates the actual workflow this paralegal cares about - the one she already sent him a stack of SOPs for, because she's that bought in.

Step two: give it to her for free. Don't charge anything. Get it implemented, get it working, let her see the time savings firsthand.

Step three: put her face on the landing page. When cold emails go out to estate planning lawyers and personal injury attorneys and healthcare defense firms - the practices that require HIPAA compliance and deal with medical data every day - and someone writes back asking if this has been done for any real law firms, the answer isn't "we're still developing it." The answer is: "Yes. Want to talk to the attorney who's using it?"

She already told him: if you get this working, I'll send people your way. That's not just a referral source. That's a sales closer. She hops on a call with a skeptical prospect, explains what her day looked like before and after, and you close deals you couldn't close with a cold email alone.

That's how you go from zero clients to ten or fifteen without spending anything on ads, without hiring a sales team, without paying a dev shop. You just needed to build the thing first - and building the thing is now a Tuesday afternoon project, not a $2,500 invoice.

The Outbound Layer: What I'd Do Right Now

Once the proof of concept is running and the landing page exists, the play is aggressive outbound. Not LinkedIn DMs. Not networking events. Cold email to Google Maps-scraped lists of law firms.

If you want to build those lists yourself, ScraperCity's Google Maps scraper pulls firm names, addresses, websites, and phone numbers from any geography you target. Pair that with an email finder to get contact-level data and you've got a targeted list of estate planning attorneys, personal injury firms, elder law practices - whatever specific pocket you're going after - without paying for a database subscription.

For the sending itself, something like Instantly or Smartlead gets you to about 20,000 emails a month at a few hundred dollars once you've got warmed domains. Buy fresh domains, give them two weeks to warm up, then start sending. If you need to move faster and want someone to manage the whole infrastructure - domains, warming, sending - there are done-for-you options that run a couple thousand dollars a month and are, frankly, worth it at that stage.

The targeting for this specific product is clear: estate planning attorneys, personal injury lawyers, medical malpractice defense firms, healthcare law, elder law - any practice that touches patient data and has HIPAA obligations. Those are the people with the pain, the compliance anxiety, and the repetitive intake workflows that eat hours every week.

If you want the cold email scripts that actually work for this kind of offer, grab the Top 5 Cold Email Scripts - there's a framework in there that maps directly to this type of niche service pitch.

What the 10-Minute Rebuild Actually Signals

Here's the broader point I want you to take from all of this.

There's a window right now where a small number of founders have quietly internalized what AI can do at the execution level - and the majority of people they're competing with haven't. The gap isn't in ideas. It's not in market knowledge. It's in the mental model of what "building something" costs in time, money, and team.

If you're still thinking about product development the way you thought about it three years ago - hire a dev team, write a spec, wait eight weeks for a build, iterate, wait again - you're operating on a cost structure that's already obsolete. Someone who has made the mental shift is going to outmaneuver you not because they're smarter or more experienced, but because they can run ten experiments in the time it takes you to scope one.

I told him: go sell this to people until they figure out they can do it themselves. Because right now, most people haven't figured it out. The window doesn't stay open forever. The guy who builds the law firm automation product and gets fifty clients in the next twelve months isn't going to be the one who spent three months waiting for a dev team to deliver v1. It's going to be the one who spent a Tuesday afternoon with Claude, got something working, gave it to a grateful paralegal for free, and started cold emailing every estate planning attorney in the state the following Monday.

That's it. That's the whole move.

I spent $160,000 learning how to build LeadShark. I rebuilt it in ten minutes. The $160,000 wasn't wasted - it's the reason I know exactly what to tell Claude to build. But the execution barrier has collapsed. The question now isn't whether you can afford to build it. The question is whether you're willing to sit down and do it.

Your Action Items If You're In This Position

If you're sitting on a service or automation product idea and you've been waiting for budget, or a dev team, or the right moment - here's the exact sequence:

• Build the landing page first. You can't do serious outbound without somewhere to send people. Claude writes clean HTML. Your wife's web business, WordPress, Go High Level, whatever - doesn't matter. Get something live that clearly explains the offer and has some social proof on it, even if it's just a description of what the product does and who it's for. Two to three hours.
• Build the proof of concept in parallel. Open Claude, go to Opus 4.5, paste in whatever documentation or SOPs your target user has given you, and ask it to build the workflow. Tell it what APIs you're connecting to. Tell it where you're hosting. Let it ask you questions. Debug with it. Four to five focused hours. Maybe a full day. You will be surprised.
• Give the first version away. Find the one person who's already bought in - the person who sent you SOPs, who said "I'll send people your way if you build this" - and implement it for free. That person is your case study, your reference call, and your most credible sales asset.
• Go hard on outbound once the landing page and proof of concept exist. Scrape your target list, warm your domains, start sending. Two weeks to warm up if you're using fresh domains. Use that time to finish the build.

If you want the full outbound system - list building, email sequencing, follow-up cadence - the Best Lead Strategy Guide has the framework I use across all of my companies.

The people who are going to win the next few years aren't the ones with the biggest budgets or the most experienced teams. They're the ones who rebuilt their mental model of what "expensive" means - and acted accordingly before everyone else caught up.

I built a $160,000 product in 10 minutes. I mentioned it as an aside. That's where I operate now.

The question is: where do you operate?