Skip to content
Studio

How the Studio operates · 4 concrete flows

Phone capture
to publish.

Four real flows. From the moment Frank taps "Drop to FrankX" on the Android share menu to the moment a published post lives on a platform. No fiction; this is what the system actually does.

~70 minutes capture-to-publish

Voice memo → five published pieces

01 — Capture (L0)

Frank walks the dog. Records an 8-minute voice memo on the OnePlus 15R about agent-framework tradeoffs.

02 — Sync (L1)

Tap "Drop to FrankX" share. Tasker classifies mime, drops to Syncthing inbox. LAN sync to Windows: ~2s.

03 — Classify (L2)

Watcher mime-sorts to ~/_inbox/voice/<date>/. SessionStart hook surfaces pending. /intake-classify dispatches audio-producer.

04 — Produce (L4)

  • ·Transcribe via Whisper (90s)
  • ·Detect 3 ideas: framework tradeoffs / personal use case / a-CES anecdote
  • ·Identify 90s podcast snippet at 03:42-05:12
  • ·Brand-voice gate runs on every text output

05 — Outputs

  • Blog draft (1500 words on framework comparison)
  • Podcast snippet (90s extracted audio + script)
  • LinkedIn carousel outline (framework matrix as 7 slides)
  • X thread (3 tweets, one per idea)
  • Newsletter section (200-word distillation)

Forcing function

NLDigital workshop in 5 days → bias LinkedIn + newsletter weights upward

06 — Operator review

Frank reviews via /admin/inbox. Edits one phrase in the carousel. Approves 4 of 5; defers podcast snippet.

~3 hours capture-to-publish

Talking-head video → multi-platform

01 — Capture (L0)

Frank records a 6-minute talking-head on iPhone (er, OnePlus). Topic: how IIS multi-agent debate catches biases.

02 — Sync (L1)

Share to "Drop to FrankX". Mime detects video → routes to talking-head subfolder. Syncthing replicates.

03 — Classify (L2)

Classifier reads frame samples + transcribes audio. Tags `talking-head-video`, spectrum=tech, archetype=Builder.

04 — Produce (L4)

  • ·Whisper transcription with timestamps
  • ·video-producer (/talking-head-ship) handles long-form composition in Remotion
  • ·content-vault-clipper extracts 3 short cuts
  • ·visual-intelligence dispatches Higgsfield Soul-ID for B-roll cutaways
  • ·hook-engineer writes 3 hook variants per short

05 — Outputs

  • YouTube long-form (8-minute final cut + thumbnail)
  • 3 YouTube Shorts (captioned, lower-thirds)
  • 3 TikTok variants (re-cut for native syntax)
  • Paired blog post (long-form companion essay)
  • LinkedIn announcement post

Forcing function

IIS OSS extraction in 4 weeks → cross-link from video description to /intelligence-system + GitHub repo

06 — Operator review

Review long-form first. Approve. Cascade approves the 3 Shorts. Schedule blog for publish day.

~10 minutes capture-to-publish

Restaurant photo + voice → IG + blog

01 — Capture (L0)

Frank takes a photo at a restaurant in Amsterdam. Records a 30-second voice note about the chef's technique.

02 — Sync (L1)

Two shares: photo + voice memo. Both land in same 10-min batch window. Watcher groups them.

03 — Classify (L2)

Classifier reads both: `photo-food` + `voice-memo` paired. EXIF GPS preserved. Spectrum=soul.

04 — Produce (L4)

  • ·food-producer reads visual: dish, composition, lighting, mood
  • ·audio-producer transcribes opinion
  • ·Pairs them: caption is sensory-specific (not generic praise)
  • ·Generates IG carousel + Threads + travel-blog fragment + Stories frame

05 — Outputs

  • Instagram carousel (4 frames: hero + 3 detail shots)
  • Threads post (single, opens conversation)
  • Travel-blog fragment (auto-appends to "Amsterdam, May 2026" entry)
  • Stories frame (lo-fi, 2-line overlay)

Forcing function

No active function for food today, but maintains soul-spectrum cadence on IG

06 — Operator review

Approve all 4. Ship to IG + Threads. Travel-blog entry queued for weekly digest.

~45 minutes capture-to-publish

Music track export → cross-promo

01 — Capture (L0)

Friday morning. Frank exports a new Suno track for the weekly album drop. Drops .mp3 into the music subfolder.

02 — Sync (L1)

Single file. Syncthing replicates. Watcher mime-sorts to `~/_inbox/music/track/<date>/`.

03 — Classify (L2)

Classifier reads metadata + duration. Tags `music-track`. Spectrum=soul. Friday cadence forcing function active.

04 — Produce (L4)

  • ·music-catalog-indexer registers the track
  • ·vis-producer dispatches NB2 album cover at 3000×3000
  • ·music-video-batch generates a lyric video draft (queued, not immediate)
  • ·content-social-distributor writes IG carousel + Spotify-companion post + Threads teaser

05 — Outputs

  • Album cover (NB2 2K+, soul spectrum)
  • Lyric video (queued for batch)
  • Instagram carousel announcement (4 frames)
  • Spotify-companion post for IG
  • Threads teaser (conversation-starter)
  • Catalog index entry (12,001st song)

Forcing function

Spotify Friday drop is the forcing function itself — this IS the deadline

06 — Operator review

Approve cover after Council 6-step gate. Ship to Distrokid + IG + Threads. Lyric video processes overnight.

The cadence this enables

Operator-time stays under 2 hours per day.

Most creator-content operators spend 4-6 hours per day on production: capture + editing + writing + posting + analytics. The Studio compresses that to ~90 minutes by moving 70%+ of the work to producers that run in parallel, gated only at the human review step.

The remaining 90 minutes is the part only Frank can do: judgment. Does this caption ring true? Does this thumbnail represent the system honestly? Does this newsletter section earn the subscribe? Producers ship candidates; Frank ships verdicts.

The forcing functions calendar provides the rhythm. Workshop in 5 days bumps LinkedIn cadence. Album drop on Friday bumps cover-gen priority. CIS MV1 keeps Bluesky alive daily. The system follows your commitments; you do not chase the system.

See the producers.

Each flow above is a composition of L4 producer specialists. Each producer has its own deep-dive at /studio/<producer>.

Producer indexForcing functions calendar