Building LaunchPad: one product brief, 42 launch channels, and the bugs that almost sank it

Short summary

LaunchPad is a free tool that takes one product brief and drafts native posts for 42 launch channels, each respecting real character limits. The AI layer runs entirely client-side via Chrome's Gemini Nano or the user's own OpenAI/Claude key, keeping marginal costs near zero. The author shares hard-won debugging lessons: TipTap v3 change-event gotchas, race conditions between server fetches and local edits, and why LLMs can't reliably count characters.

• •One product brief generates channel-native posts for 42 platforms with enforced character limits
• •AI runs on-device via Gemini Nano or user-supplied API keys — zero server-side AI cost
• •Real debugging lessons: TipTap event handling, state race conditions, LLM character-count failures

Generated with AI, which can make mistakes.