Studio · v1.0.0 · 2026-05-13
The Studio is the operating system behind frankx.ai content. Eight L4 producer specialists. Fourteen platform personas. Three operator archetypes. One Android-native inbox. Phone-to-publish in under two hours.
The single-capture pipeline
Capture lands on the phone. Syncthing replicates to the machine. The classifier reads multimodally. The orchestrator dispatches producers in parallel. Operator approves once, ships everywhere.
L0 CAPTURE OnePlus 15R + Android phones — voice, video, photo, screen L1 SYNC Syncthing-Android → ~/_inbox/ (LAN <2s) L2 CLASSIFY content-intake-classifier — multimodal pass per file L3 ORCHESTRATE multimodal-orchestrator — parallel L4 dispatch L4 PRODUCE 8 specialists (vis, video, audio, music, prose, screen, food, travel) L5 OPERATE /studio (public) · /admin/inbox (private, W22) L6 DISTRIBUTE Native per-platform publishing (manual until volume>200/mo) L7 LEARN hook-learn (live) · post-time-learn · thumbnail-learn (W22)
Full architecture in docs/superpowers/specs/2026-05-13-content-ops-architecture.md. Setup guide for the L0/L1 Android side: docs/ops/ANDROID-INTAKE-SETUP.md.
The L4 specialists
Each specialist reads from the same persona matrix and capture-type registry. Adding a new producer follows the VIS shipping pattern — typed substrate, skill, command, agent, public page. About a day of work per specialist.
Live. Drop images, get a per-platform strategy.
/visual-strategyLive (existing /talking-head-ship). Substrate integration W20.
/talking-head-shipLive. Voice memo → blog + podcast + newsletter excerpt.
/audio-produce60% live (existing music-video-batch + Suno pipeline). Substrate integration W20.
/create-musicLive. Text seed → cross-platform with brand-voice gate.
/prose-produceLive. Screen-record → tutorial + README + Short. SE-specific keystone.
/screen-produceLive. Restaurant photo + optional voice note → IG + Threads + travel-blog.
/food-produceLive. Travel photo with GPS → IG carousel + blog + Stories.
/travel-produceThree operator archetypes
Almost nobody operates all three archetypes credibly. Most creators are creator-only; most influencers shed technical depth as they grow; most solution engineers don't ship public content. The combo is the moat.
Creator · 5 platforms
Music + watch + workshops. Spotify drops, IG, YouTube long-form + Shorts, TikTok, Stories.
Influencer · 7 platforms
The AI Architect personal brand. LinkedIn, X, Threads, Newsletter, Bluesky, Podcast.
Solution Engineer · 2 platforms
The Oracle EMEA bridge story. GitHub, Newsletter, Podcast, LinkedIn, YouTube long-form.
13 capture types
Voice memo on a walk. Restaurant photo. Talking-head from the studio. Screen-record of a system running. Architecture sketch from Excalidraw. The classifier reads all of them.
voice-memo
Voice memo
Phone-recorded voice note. 1–15 min typical. Source for blog posts, podcast snippets, newsletter excerpts, Bluesky publishes.
talking-head-video
Talking-head video
Phone or camera capture of Frank speaking on camera. 1–10 min typical. Source for YouTube long-form + Shorts + TikTok + Reels.
b-roll-video
B-roll video
Silent or ambient capture (no on-camera speaking). 5–60 sec typical. Used as cutaways in longer compositions.
screen-record
Screen recording
Phone or desktop screen capture. 30 sec–10 min. Source for tutorial blogs, GitHub README sections, /watch/shorts dev content.
music-track
Music track
Suno export OR final mix from another DAW. 2–5 min typical. Source for Spotify drops, IG promo, music video pipeline.
music-seed
Music seed
Phone field recording, hummed melody, found-sound clip, loop idea. 5–30 sec typical. Source for Suno prompts.
photo-hero
Hero photo
Deliberate composition meant as a hero image. Differentiated from photo-utility by intent + brand-fit signal.
photo-food
Food photo
Restaurant or home-cooked food capture. Pairs with voice-memo for opinion. Source for IG, Threads, travel-blog.
photo-travel
Travel photo
Location-tagged story shot. EXIF GPS expected. Source for IG carousel, blog entry, Stories.
photo-utility
Utility photo
Screenshot, reference photo, document scan, low-effort capture. Source for blog inline imagery, slack-style sharing.
document
Document
PDF, research note, transcript, dictation, brain dump. Source for blog drafts, newsletter excerpts, case studies.
quote
Quote
Highlighted text from a book, podcast clip, conversation snippet. Short. Source for X posts, quote cards, library annotations.
architecture-snap
Architecture diagram
Excalidraw, Whimsical, draw.io export, or hand-drawn sketch of a system architecture. Source for LinkedIn carousel, GitHub README, blog post diagram.
The runtime
Production-shipped today. L4 producer specialists ship W20–W21 on the same pattern.
Command
/intake-classify
Classifies pending captures. Walks the 5-phase run protocol. Writes per-batch manifest + operator summary.
Command
/intake-watch
Start, stop, or check the watcher daemon. The L1→L2 boundary that mime-sorts Syncthing drops.
Agent
content-intake-classifier
Reads every new file multimodally. Tags type, spectrum, brand-fit, conversion-potential. Suggests producer dispatches.
Agent
multimodal-orchestrator
Fans out classified batches to L4 producers in parallel. Collates outputs for operator review.
Daemon
scripts/intake-watcher.mjs
Node fs.watch process. Watches ~/_inbox/dropped/, mime-sorts to typed subfolders, updates pending queue.
Init
scripts/init-inbox.mjs
One-time setup. Creates ~/_inbox/ and ~/_archive/ filesystem per the canonical schema. Idempotent.
Substrate, runtime, public face. Spec at docs/superpowers/specs/2026-05-13-content-ops-architecture.md.