The model already found titles that score higher. Try one.
Hetzner box, 92 Telegram RPC methods via MCP, dev-branch self-edits with a daily safety-gated merge to main. What broke, what worked, real costs.
ForesynWanna keep in touch?
Built this solo over a weekend. Soft-launching before the HN post on Monday. If you scored a draft and the prediction either nailed it or whiffed, I want to know.