How Pikari's startup ideas are made

New ideas are generated every day with Claude, Anthropic's AI model. Each batch is spread across categories, industries, budgets, and timelines, so it doesn't collapse into a dozen versions of the same AI app, and the model is asked for ideas a solo founder could realistically bootstrap: specific, not generic.

Every idea is then checked automatically before it reaches the catalog. Ideas with missing or invalid parts are rejected and regenerated, and anything that duplicates or nearly duplicates an existing idea is dropped. No person reviews each idea before it appears in your feed; we'd rather say that plainly than pretend otherwise.

• Who it's for and what it does. A one- or two-sentence pitch that names the audience first.
• Budget, time, and skills. Roughly what it costs to start, how long a first version takes, and which skills it needs, so ideas can be matched to you.
• Ways to make money. One to three concrete monetization paths, not “grow then figure it out”.
• A 3-step MVP roadmap. How to validate demand, what to build first, and where the first customers come from.
• Why now. The trend signals that make the idea timely.

Each idea carries a validation score out of 100. It's the AI's estimate, made when the idea is generated, of how promising it looks: likely demand weighed against competition and how feasible it is for a solo founder. It is not measured from market or search data, and it isn't a prediction. Treat a high score as “worth a validation weekend”, and let real customers give you the real answer.

Onboarding asks about your interests, skills, available budget, and the time you can realistically commit. Those answers score every idea, so a solo builder shipping with AI tools on weekends sees different ideas than a funded team that's hiring.

Then the feed learns from your ratings. Ideas similar to ones you love move up; categories and industries you keep passing on move down. It also learns from founders who rate the way you do, so ideas they liked can surface for you.

Ratings shape what gets made, too: when many people pass on a kind of idea, and say why, future batches lean away from it.

The ideas on /try come from a separate, fixed pool, made the same way and generated in advance so every combination of answers has ideas that fit. Nothing is generated live while you click, and the same answers always give the same deck.

An idea from Pikari is a hypothesis worth testing, not a validated business. The score, the monetization paths, and the roadmap exist to make your own test faster, not to promise an outcome.

The blog and the idea hubs follow the same approach: practical, specific, and honest about what validation actually takes.