Launch Sphere

Launch Sphere is a model-neutral prompt engineering product: a guest can land on Prompt Enhancer, rewrite a rough idea into four styles, and leave without an account. The product bet is the library — prompts that run across GPT, Claude, and Gemini rather than sitting in a vendor silo.

I built it as a shipped app, not a case-study deck. That meant the privacy questions were real: user prompts are sensitive, guests still need a usage cap, and billing cannot become a second data store.

What I put in the data story

• Prompts are processed to run the tools and stored only so people can get them back. They are not sold and not used to train models.
• Cards stay with Stripe. Image files sit on Cloudflare R2. Models are routed through OpenRouter. Hosting and the database are named in the public policy — no mystery processors.
• Guest trials are capped with essential cookies, not an ad pixel. There is no third-party advertising or tracking cookie.
• Library entries can be deleted in-product. Account export or erasure goes to support.

The live policy is at launch-sphere.com/privacy. The interesting part is not the page — it is that the product had to stay usable with that policy: three free tries, no card, no dark-pattern consent banner.