Notes6 min read
Everyone's vibe-coding. Nobody's getting customers.
Usually because nobody planned how to reach them. Most AI build courses stop at deployed. Positioning, a launch plan and real conversations with customers are the missing half, and you can learn them.
You've probably seen this story, or lived it. Someone spends a few weekends building an app with an AI agent. It works. It looks good. They post it, tell some friends, and share it in a community or two. About twenty people visit. A couple sign up. Then nothing happens.
The app wasn't the problem. Nobody had decided who it was for, how those people would hear about it, or what would make them switch from whatever they use today. The build got all the attention. The selling got none. That's why so many AI-built apps get no users.
Why AI-built apps get no users: building stopped being the hard part
"Vibe coding" means describing what you want in plain words and letting an AI agent write the code. It works well enough that the number of new apps has exploded. Lovable, one of the popular app builders, said in December 2025 that more than 100,000 new projects were being built on its platform every day.
That's good news for anyone with an idea. It also means your app is launching into a crowd. When making something is cheap, the scarce things are attention and trust. A working app is now the starting line, not the finish. We wrote about why judgment is the scarce part now.
The best data we have on why products fail points the same way, with a caveat. In March 2026, CB Insights looked at 431 venture-backed startups that shut down since 2023. Running out of money showed up in 70% of cases, but the report treats that as the final symptom. Poor product-market fit, meaning not enough people wanted the product enough to pay, appeared in 43%. Those were funded companies with teams, not weekend apps. Still, the lesson carries over: products rarely die because the code was bad. They die because the market never showed up.
Why build courses stop at "deployed"
Most courses that teach non-developers to build with AI end at the same place. Your app is live on a custom domain, maybe with a database and payments. Then the course is over.
There are fair reasons for that:
- "Deployed" is a clean finish line. You can see it and check it. "Has paying customers" is messy and slow.
- Selling is uncomfortable. It means talking to strangers and hearing no. Paul Graham noticed this years ago. Founders would rather stay home and build than go out and talk to people who will "probably be rejected by most of them," as he put it.
- Building feels like progress. Each new feature is visible. Each unanswered message feels like failure, so it's tempting to go back to features.
The result is a gap. You learn to make the thing, then you're on your own for the part that decides whether it matters.
What the missing half looks like
Selling isn't a talent you either have or don't. It's a set of small, repeatable jobs. Agents can do much of the legwork, as long as you make the decisions and approve anything that goes out under your name.
Here's how that plays out in the course, using a made-up example. Kudos is a small app that collects and displays customer testimonials. In the course you work on your own project.
| Lesson | What you decide | What you leave with for Kudos |
|---|---|---|
| Positioning | Who it's for, what they use today, why you win | "For freelance designers who lose good client feedback in email threads" |
| Launch plan | Where your people already gather, in what order | A launch sequence for two or three communities, with assets ready |
| Prospect research | Which 50 named people fit the profile | A list researched in parallel by agents, checked by you |
| Outreach with approval | What to say and who gets it | Personal drafts that a hook checks for tone; nothing sends without you |
| First customer calls | What you need to learn | A call brief before each call, a follow-up after |
| Real testimonials, AI-produced | What proof to show | Real customer quotes turned into clips and quote cards |
Then the Run stage keeps it going without you watching every hour. A launch-day watch checks signups, errors and replies on repeat. A Monday digest sums up signups, revenue and churn. A daily routine sorts new leads and feedback into the right place. These are small automations, but they're what keep a launch from turning into silence a week later.
Positioning comes first
If you do one thing after reading this, write a positioning sentence. Positioning is the short answer to "who is this for, and why would they pick it?" It shapes everything that follows: your landing page, who you contact, and what you say on calls. Before that, it helps to know the idea is worth building at all, and writing the press release first is a cheap way to find out.
A simple shape that works:
For [a specific kind of person] who [has this problem], [your product] is [a kind of product] that [main benefit]. Unlike [what they use today], it [the one difference that matters].
For Kudos, a first draft might read: "For freelance designers who lose good client feedback in email threads, Kudos is a testimonial collector that turns a client's kind words into a wall you can embed on your site. Unlike screenshots of emails, it asks at the right moment and gets the client's approval on the final quote."
It won't be right the first time. That's fine. The point is to have something specific enough to be wrong, so real conversations can correct it.
Try it
Give yourself one hour, before you add another feature.
- Write your positioning sentence using the shape above. Ask Claude to poke holes in it, then rewrite it yourself.
- Name twenty real people who fit it. Not "designers on LinkedIn," but actual names. If you can't find twenty, that's useful information about your market.
- Message five of them by hand. Don't pitch. Ask how they handle the problem today and whether you can watch them do it. Graham's advice still holds: the most common early job that doesn't scale is recruiting users one by one.
- Write down what surprised you. That note is worth more than your next feature.
If an agent drafts those messages for you, read every one before it goes out. Your name is on it, and your first customers will remember how you treated them.
The course spans Sell and Run on purpose
Chief of Agents follows one product from an idea to a launched product with real customers. That's why selling and running get two full stages instead of a bonus video at the end. Building is where most courses stop. In ours, it's the middle. You can see the full journey on the course page.
Sources
Learn this inside Claude Code.
Join the waitlist for Chief of Agents. You'll get the free beginner's handbook and skills right away, and hear first when the course opens.
Keep reading
6 min read
From designer to builder: directing a crew of agents at work
What working with AI agents looks like day to day after ten years in design: the work of direction, and what stayed the same.
6 min read
Building with AI got cheap. Judgment didn't.
When AI can build an app in an afternoon, the hard part is judgment: what to build, what good looks like, and who will pay.
6 min read
AI closed the skill gap. It widened the taste gap.
Why does AI output look generic? It drifts to the average. Here's why, plus a side-by-side drill that trains your eye to fix it.