Build a waitlist page
One sentence and one email field, to test whether anyone wants the thing.
A waitlist page is a cheap test of a pitch. You write one sentence and add one email field, and the answer comes back as a number. The page's only job is to make that sentence land.
A static page can't store emails and a waitlist doesn't deserve a backend. Pick a form service below (or a small Cloudflare Worker if you want to own the list) and the prompts adjust.
The spec and the cuts
- Above the fold: the sentence, one line of proof that you exist, and the email field.
- One file under 50KB, since people open these links on phones and leave if it's slow.
- Say what you'll send and how often ("one email at launch, maybe two before").
- Cut: analytics, chat widgets and cookie banners.
The prompts
Pick where the emails go and what else you want, and the prompts change to match. Paste them into your agent in order. Prompt 2 reads the page the way a stranger would and fixes what it finds.
The copy buttons grab the current version.
1 · Build it
2 · The stranger test
3 · Ship it
4 · Test the pitch on humans before the internet
Where agents go wrong
- Manufactured urgency. "Only 17 spots left!" is slop, and smart visitors spot it. Show numbers only if you have them.
- Feature grids for vapor. Every section below the fold is a reason not to sign up.
- Validation theater. Skip the email regex.
type=email requiredis enough. - Forgetting the unfurl. Most people see this link as a preview card in a feed or DM. Without OG tags it shows up as a blank box, and fewer people click.
- Databases. If your agent starts a schema for a one-field form, stop it.
Why prompt 4 is in the plan
You can't judge your own pitch, and people who like you lie about signing up. Reviewers can tell you what they think the product is and which word lost them. Prompt 4 connects your agent to my MCP server so two or three people can pin notes on the live page before you post it.