Playbooks
Build what the big AI companies can't copy
A playbook for AI products that need to stay useful when a new release from the AI company gives customers the same basic results in its own app.
The hard part: spotting which parts of your product a platform vendor could ship next, and building around the workflow instead.
Who pays
The hard part of the business is giving people a reason to use your product beyond the model's raw output: a workflow, your own data, or distribution that the chatbots cannot copy.
How to build it
See what chatbots already do
Before you build, give the exact job to ChatGPT, Gemini, Claude, and Grok. If they already do it well, plan to sell more than that output.
Name what chatbots can't copy
Write down what you would add that ChatGPT, Gemini, Claude, and Grok can't copy: a workflow, your own data, or distribution. If you can't name anything yet, keep working on the idea until you can.
Recheck after each model release
After each major model release, run the same test again in ChatGPT, Gemini, Claude, and Grok. This shows you any new overlap early, before it shows up in your numbers.
Build your own email list
Set up an email list, with Supabase storing signups and Resend sending email. This gives you a channel to your audience that doesn't depend only on search or on one model.
Sources on X
- levelsio · Jun 30, 2026
The founder of Photo AI says revenue crashed at most competing products after Nano Banana Pro launched. By their own estimate, about 30% of users switched to Gemini and similar apps to take AI pictures. They argue that if you don't control the models you don't have much moat, and that training a model like that costs $100M.
- danielkbuilds · Sep 3, 2026
The poster says their international tax calculator site, grown only through organic SEO, made $1,000 in ad revenue in the last 30 days. They also say that as more people use AI, sites like this will probably die out over the long term.
Ideas are curated from public posts and untested by us. Read the full note.