Skip to main content

Thruit — Tester Brief

Thruit is a living knowledge graph for the people, systems, and recurring themes around your work. You record or paste meeting content; Thruit auto-maintains a structured set of markdown files about everyone you mentioned, every theme that's come up, and every commitment that got made. You don't curate it — it curates itself.

Right now Thruit fits the solo knowledge worker embedded in one organization — someone like a director, manager, consultant, or operations lead trying to keep mental track of dozens of people and overlapping initiatives.

What you can do today

  • Sign up with email + magic link
  • Onboard by telling Thruit your role, your company, and 3–5 themes you want it to track for you
  • Record a meeting in-browser via your microphone (works on phone or laptop), or paste in notes / transcripts you already have
  • Watch the live pipeline indicator as your recording is transcribed and structured
  • Browse your workspace's file tree: people/, themes/, transcripts/, prep/, plus auto-generated rollups like pain-points.md and opportunities.md
  • Click any fact to jump to the exact line in the original transcript where it was said — this is the trust feature, every claim is traceable
  • Pre-meeting prep documents (e.g. prep/<person>-<date>.md) get drafted automatically a day or two before scheduled meetings
  • Switch between multiple workspaces if you wear multiple hats (employee + consultant, etc.)
  • Power users: git clone your entire workspace and edit it locally in Obsidian, Cursor, Claude Code, or any markdown editor

Why everything lives in git (this is the unusual part)

Most knowledge tools lock your data inside their database. Thruit doesn't. Your workspace is a private git repository that you can clone, browse, edit, and own forever. That gives you:

  • No vendor lock-in. It's plain markdown in plain folders. If we disappear tomorrow, you walk away with git clone.
  • Complete history. Every fact added, every edit, every auto-update is a git commit. Nothing ever silently changes — you can see exactly when something was added and from which conversation it came. Hand-edits and machine-edits sit side-by-side in the same history.
  • Source-traceability on every fact. Click a bullet and see which transcript line it came from. This isn't possible in tools that store knowledge as opaque database rows.
  • Use the editor you already love. Pull the repo into Obsidian for graph view, Cursor or Claude Code for AI-assisted refinement, or any markdown editor you prefer. The web app is one way to interact with it; it's not the only way.
  • Branching workflows. Want to draft a major reorg of how you've categorized something? Branch, edit, merge. Just like code.

The product is the auto-maintenance. The data is yours.

Step-by-step testing path

  1. Sign up at the URL Jeremy sent you. Use a real email — the magic-link flow needs to deliver to you.
  2. Onboard. Three quick prompts: your role, your company, and themes you want tracked. Be honest — Thruit will use these to bias what it pulls out of recordings.
  3. Record a meeting at /record. ~30–60 seconds is plenty for a first test. Mention 2–3 people by name, an opinion they hold, a commitment, a recurring theme. You can leave the title field empty — Thruit will name it for you.
  4. Watch the pipeline. The page will show 4 steps as your recording is processed. Takes about 2 minutes end-to-end. Stay on the page or come back to it.
  5. When pipeline says "Done", click "Go to workspace" to land in your file tree.
  6. Open the people/ folder. You should see a markdown file for each person you mentioned, populated with the facts you said about them.
  7. Click any fact — you should jump to the exact line of the transcript where it was said.
  8. Wait ~15 minutes, then refresh that person's file. You should see it has been re-organized into sections (Background, What they want from you, etc.) automatically.
  9. Try /paste if you have meeting notes already in a doc somewhere. Same pipeline, no recording needed.
  10. Come back tomorrow morning. Overnight, Thruit will have generated pain-points.md, opportunities.md, and a cross-cutting patterns.md based on everything you've fed it. Check those.
  11. (Optional, for power users): ask Jeremy for the git clone URL. Pull your workspace locally and open it in Obsidian or your editor of choice. You're looking at the same data the web app shows, plus full git history.

What we want from you

  • Friction reports. Anywhere it felt slow, confusing, or scary.
  • Trust failures. Any extracted fact that's wrong, attributed to the wrong person, or missing the source link.
  • Missing affordances. Things you wanted to do that the UI didn't let you.
  • Anything that surprised you — good or bad.

It's an early system. Things will break. Tell us when they do.