Founder field notebook
01 TODAY
See what is current, what needs your call, and which proof is ready to carry forward.
Every saved artifact should point to one clear next move instead of leaving the work inside a conversation.
Loading the latest saved artifact.
One map for where the factory is active, proven, reusable, or waiting on a founder decision.
Import a VoiceSet output or create the first decision manually.
02 PROJECTS
The project layer keeps artifacts grouped as a living body of work. No project state is hidden inside a model session.
saved artifacts
active modes
most recent update
Projects are the durable containers for decisions, build plans, reviews, and handoffs.
The portfolio is loading from the local artifact store.
03 WORK
Start from a voice transcript, chat, agent result, or Markdown artifact. The original source stays separate from anything VoiceSet derives from it.
Paste the provider transcript exactly as received. VoiceSet stores it as a source record. Additions from chat and later summaries are separate records, so the original voice context is never silently replaced.
No source session has been saved in this form.
Use the same text whether you copy it from VoiceSet or save it as an .md file. Studio reads the file locally and does not save anything until you review the form below.
Nothing has been imported or saved.
This removes only the selected Studio artifact. Download a backup first if you may need it again.
04 REVIEW
Choose the artifact and mode for the next session. Studio keeps the accepted state and gives you a compact handoff command.
Review is the bridge between stored proof and the next useful conversation.
When the VoiceSet app is connected to your personal ChatGPT account, this short command makes ChatGPT load the full artifact, context, and standards through Studio.
Use the pre-flight before voice. A connected ChatGPT app can save the finished artifact directly; the text save command remains the fallback.
Studio records event names and references, never conversation content. Resume success and failure are explicit because generating a command does not prove the next system used it.
No tracked workflow is active.
Give an agent one goal, one expected output, and explicit limits. Studio prepares the handoff only after you approve it. Results return as linked work and never overwrite the source artifact.
No agent task is active.
05 TIMELINE
Artifacts, voice sessions, agent work, workflow events, and your own dated notes in one local timeline.
06 CONNECT
Use the VoiceSet app in ChatGPT so conversations can load context and save artifacts directly without moving the content by hand.
Connector health and bearer authentication.
Private source handling and artifact lineage.
Human approval before anything is sent externally.
The connected command refuses to expose Studio when the key is missing or weak, Tailscale is disconnected, the server is unhealthy, or port 4177 is running without bearer authentication.
VOICESET_OAUTH_OWNER_SECRET in .env.npm run connected:start/mcp URL shown by the connected command.Your ChatGPT subscription signs you into ChatGPT. VoiceSet separately authorizes that ChatGPT connection through OAuth; Studio never receives your ChatGPT password or private chat history.
http://127.0.0.1:4177/api/healthDownload one JSON file containing every artifact and the content-free local event ledger. Nothing is deleted or moved.
07 ACTIVITY
An append-only record of workflow signals. It contains no transcript or automatic memory claim.
explicit workflow IDs
local JSONL records
optional average / 5
Newest first. The source file is data/events.jsonl.