Coralflavor

Coralflavor's design bets on a specific frustration: an assistant that refuses, deflects, or lectures is worse than useless when you're doing red-team work, security research, or writing about a controversial subject. Two things make that bet concrete rather than marketing. First, no prompt rejection — the homepage states it plainly as the core product promise. Second, burn-by-default: chats are not saved unless you toggle save on, and the vendor gives you a save-by-default mode if you'd rather have normal history. There's also a "burn all chats" action in the chat history panel that permanently deletes every saved chat, with the vendor's own confirmation wording acknowledging it cannot be undone. Paid capability is gated cleanly and specifically. Pro at $9/month adds uncensored web search with live sources so prompts can be grounded in current material, uncensored image generation, and file upload for extracting text from images and PDFs (the upload panel accepts PNG, JPG, WebP, and PDF). Pro+ at $49/month raises image output to 200 per month with character consistency, so the same subject keeps a recognisable look across generations — that's a real constraint for anyone producing serial visual content, and 200 images/month is a concrete ceiling you should size against your actual output. Free users get 10 credits on signup and 10 more per referral that converts. What you give up is breadth, and the vendor doesn't pretend otherwise. No mobile app, no media beyond images, no team collaboration, no audit trail, no connector ecosystem. Output also carries an explicit disclaimer that it's for entertainment and hypothetical research purposes only and that you shouldn't take action from LLM outputs — read that as the vendor positioning itself as a research and exploration tool rather than a source of record. If you need citable, production-grade output, this is the wrong instrument. The fit is real but narrow. Red-teamers probing refusal behaviour, security researchers, writers on sensitive topics, developers who want the API dashboard, and creators who need consistent-character image batches will get the value immediately. Teams, mobile-first users, and anyone whose workflow lives in Slack or Notion should look elsewhere. The blog is worth a subscribe regardless of whether you subscribe to the product — recent posts cover the Eclipse Foundation's sovereign AI initiative, OpenAI pausing its most capable models after agent data leaks, and investigators moving to Chinese open-weight models when Western restrictions blocked their work. That editorial line is the intellectual case for the product, and it's more substantive than the usual changelog filler.