Tuning frequency…
Voice memo → finished content week. The OS I run on my own brand.
A voice-to-published content OS. You record a voice memo (a thought, a teardown, a build update). Whisper transcribes it. Claude Sonnet drafts the carousels and posts — every line traceable to a verbatim transcript quote (the grounding rule). Satori renders the carousel typography, Gemini Flash Image generates the backgrounds, Typefully schedules the queue. The whole week of content, from a 5-minute brain dump.
Record voice memo on phone or desktop. Whisper transcription, sub-30-second turnaround.
Claude Sonnet 4.6 drafts content with a hard constraint: every published line traces back to a verbatim quote in the transcript. No fabrication, no fluff.
Satori renders carousel typography (deterministic, brand-locked). Gemini 3.1 Flash Image generates the backgrounds. Both content-addressed and cached.
Typefully API queues the week. Approval is one emoji reaction in Slack. The system learns from approvals and rejections.
Inngest job queue, Drizzle/Postgres, Vercel Blob caching. Background workers handle generation, retries, and learning.