What Transcripted is actually for
Dan Shipper from every.to just shipped something called Proof. It’s a free, open-source document editor built from the ground up for agents and humans to work in the same doc together. Not an AI assistant bolted onto a doc editor. A doc editor built assuming the agent is a first-class participant.
That framing stopped me.
Because I realized I’ve been building Transcripted the same wrong way everyone else is — optimizing for the human reading the transcript. That’s the wrong user.
The right user is your agent.
When you have a personal agent — and everyone will, this isn’t a distant future — it needs to know your world. Who you talk to. What you decided. What you promised. What’s been on your mind for the last six months.
That all lives in your meetings. And right now it’s completely dark to any agent you’re using because nobody built the piece that captures it and makes it usable.
That’s what Transcripted is. Not a transcription app. The context layer for your personal agent.
Every meeting gets saved as a clean markdown file on your Mac. Speaker-labeled, timestamped. The voice fingerprinting means it knows Sarah across 50 meetings, not just this one. That file is readable by any agent that can see your filesystem — Claude Desktop, any MCP-compatible tool, whatever you’re running.
Your agent now knows who you’ve been talking to. What you decided. What you owe people. That’s not a transcript. That’s memory.
Proof is built agent-first. The whole product is designed around the assumption that agents are participants, not afterthoughts.
I need to build Transcripted the same way.
Not MCP servers, not integrations, not a new chat interface. Just the simplest possible thing: put your conversations in a folder, make the files rich enough that any agent can use them, and get out of the way.
The red bar: I shipped the right technology with the wrong intent. The diarization, the voice fingerprinting, the local pipeline — all of that is correct. But I was optimizing for “good transcript” when the thing that actually matters is “context an agent can use.”
Same app. Different north star. That’s the shift.