I Accidentally Launched My Course. Then I Fixed the Waitlist Live On Air in 45 Minutes.

The course is called AI Builders. It opens in May, five weeks of live building with a cohort of non-technical founders. I've been putting off the waitlist page for weeks because I didn't have time to "do it properly."

Then I put it in the newsletter anyway, because my logic was "the course is real, I should mention it." Classic founder self-sabotage: you half-ship the story and then scramble to make the reality match.

Problem: the newsletter goes out wide. People read it, clicked through, found a broken page, and some of them paid anyway. That is both flattering and terrifying. Flattering because people wanted in enough to pay a stranger. Terrifying because the payment flow hadn't been touched.

So the pressure became real. Not "maybe I should fix this next week." More like "fix this in the next hour before a hundred more people see it."

Going live on Build Hour was the forcing function. I had 45 minutes left in the session and a landing page that looked like I'd written it at 2am on a phone. Let's go.

The entire build started with one prompt:

"Make this an incredible experience."

That's it. I pointed Claude at the existing landing page directory, told it this was the course website, and asked for a plan. No feature list. No wireframes. No "here are 12 references I like." Just the outcome I wanted.

Then I added one more line that changed everything:

"Ask me any questions you need with question and answer."

This is the trick that costs me 10x fewer tokens than how most founders work with Claude. Let me unpack it.

The default behaviour when founders work with Claude Code goes like this:

• Think about the task for 10 minutes
• Write a 400-word prompt stuffing every preference, constraint, and reference you can remember
• Paste it in
• Get a response that misses three things you forgot to mention
• Reprompt

That is a waste of both your brain and your tokens. Half of what you typed wasn't needed. The other half missed what actually mattered.

The inversion: give Claude the goal, then let it interview you.

When you end a prompt with *"ask me any questions you need, with question and answer"*, Claude shifts from executor to product manager. It looks at the problem, figures out what information is missing, and asks you specifically for that. You answer in bullets. You spend maybe 2 minutes on the interview. Claude has what it needs to start building.

Why this is better:

1. You don't pre-guess what matters. Claude decides what it actually needs.
2. You answer in short bursts. Two words beats two paragraphs.
3. It catches assumptions. The questions often reveal that you haven't decided something yet.
4. Token cost is tiny. Short questions, short answers, cheap cache.

On the course page, Claude asked me four things upfront: What's the launch strategy, waitlist first or direct? Cohort size cap? Testimonials available? Which viral mechanics do you want? I answered in three sentences. Then it built the page.

Think of it as outsourcing the "what am I actually trying to do" step to the model instead of grinding it out yourself. Works for landing pages, PRDs, refactors, pitch decks. Anywhere the brief is fuzzy.

Here's the table the page ended up with, straight out of the 45-minute build:

I love this structure. It rewards generosity but stacks. If you bring 5, you've essentially turned the cohort into a mini-community before it starts, and your seat is on the house.

Claude also proposed:

• A gamified waitlist with point totals and public leaderboards
• Pre-paid early-bird pricing (deposit to lock a seat)
• "Limit the spots" scarcity framing ("50 seats only")
• FAQ section, credibility bar, hero headline variations

What I kept: the friend-referral tiers, the early-bird pre-pay option, the "50 seats only" language, the FAQ.

The pre-pay early-bird option is worth calling out. Deposit to lock a seat, get a discount. It does three things at once: filters for founders who are genuinely ready to commit, gives me real revenue to plan capacity around, and removes the "I'll think about it" state that kills most waitlists. A signed-up waitlist signal means nothing. A $99 deposit means they're in.

What I killed: public leaderboards. They'd create the wrong energy for a learning cohort. You don't want founders racing to rack up referral points when the whole community is supposed to be about collaboration. Leaderboards are for hustle products. This is a small, warm, weird little pod of people learning to build. Different vibe entirely.

Let me be honest about something that gets glossed over in "how I launched my course" posts.

The "50 seats only" framing on the page? Claude suggested it. I approved it. It's marketing scarcity, not a genuine capacity constraint. I don't yet know if I can handle 100 people or 30. The honest answer is "probably somewhere between 30 and 80, depending on how much time I spend in the community."

Scarcity that isn't real is marketing. Scarcity that is real is operational. Both can be on the same page, but be clear with yourself which is which.

So why keep it? Because a number beats no number. "50 seats" forces a decision. It's also a promise to future-me: build the operational capacity for 50 or drop the number.

The trap most founders fall into: they either skip scarcity entirely (because "I don't want to seem pushy") or they copy-paste a number they saw on someone else's page (because Claude suggested it). Neither works. The middle path: pick a number you're prepared to deliver on, and treat it as a commitment, not a ceiling.

Here's the blunt truth about launch pages: most of them read like marketing slop. Generic lines that could live on a billion other websites. "Transform your business." "Unlock your potential." "Join the revolution."

No one cares. You've said nothing.

When I went through the draft page, the stuff I cut was:

• Generic outcome statements that didn't say *what exactly* changes
• Testimonial placeholders (I don't have real ones yet and fake ones kill trust instantly)
• A "learn AI" headline that told you nothing about the specific shift this course creates
• Bloated FAQ covering questions nobody asks

What I kept: specific customer language, clear outcomes ("you'll walk away with a working project"), who it's for ("non-technical builders") and explicitly who it's *not* for, the friend-referral mechanic, and the pre-pay option.

The rule: if a sentence could appear on a competitor's site without changing a word, delete it. Every line of copy has to be specific enough that nobody else could claim it without lying.

A founder asked me in the Build Hour chat: "If I had 45 minutes and wanted to launch something today, what should I do?"

My answer was three words: just do it.

The mistake I kept making for weeks was waiting for the page to be right before I mentioned the course. The mistake that actually worked was the opposite: mention it, let the feedback force the build.

Embarrassing version of a live product beats perfect private version every time. People paid me for a broken page. They'll pay you for yours.