Developer builds AI voice filter that rejects its own bot's clichéd headlines

On April 30, 2026, a self-taught developer in Sacramento reached a milestone when his publishing bot automatically rejected an AI-generated headline containing marketing clichés and successfully rewrote it on the second attempt. The developer, roughly 60 days into learning software development, built a filter chain that scans candidate titles against a banned-phrase list, blocked title formats, and a soft-signal score measuring sterile or hedging language. If any check fails, the bot re-prompts itself with the specific rejection reason attached and logs each retry attempt for later review. The project was motivated by the developer's need to maintain a consistent personal voice across both his content and paid API endpoints, arguing that voice consistency directly drives customer retention. Building and tuning the filter took approximately six hours, requiring iterative adjustments after an overly strict initial threshold sent the bot into unresolvable retry loops.