Conversations with receipts.

What should an AI chat reveal before I trust it?

Not a trust badge. Evidence: the exact model and resolved provider, the context that was sent or removed, tool and source provenance, token usage, latency, cost, and every branch or transformation that produced the answer.

{
  "manifest": {
    "author": "Flock Capital, LLC",
    "inference": false,
    "releaseStage": "beta",
    "source": "founder-authored launch replay"
  },
  "usage": {
    "inputTokens": 0,
    "outputTokens": 0
  }
}

And if I disagree with the answer, or decide to leave?

Inspect the context, fork the turn, compare regenerations, merge the useful parts, export the whole ledger, or bring your own key. Causerie does not train on your content, hide a route, or turn your history into an exit fee.

{
  "manifest": {
    "author": "Flock Capital, LLC",
    "inference": false,
    "releaseStage": "beta",
    "source": "founder-authored launch replay"
  },
  "usage": {
    "inputTokens": 0,
    "outputTokens": 0
  }
}

So what exactly is launching?

Causerie is in open beta and this launch replay is a release candidate. The product is live, but GA remains gated on the recorded reliability window, external review, outside acceptance, and counsel. This page will say generally available only after those receipts exist.

{
  "manifest": {
    "author": "Flock Capital, LLC",
    "inference": false,
    "releaseStage": "beta",
    "source": "founder-authored launch replay"
  },
  "usage": {
    "inputTokens": 0,
    "outputTokens": 0
  }
}