firehose> #llmops

Stakeholder Clone

Reconstruct a specific real person's judgment from their artifacts, then consult that clone to preview how they will react to your work — before you submit it. The source calls it "AI time travel": build a system that gives you the feedback the real stakeholder would give, ahead of the finalized version, so you can act on it while there is still time to change the work. You clone the person by having the model ingest their available corpus — public content (blog posts, talks, podcasts) or private artifacts you legitimately hold (a manager's Slack messages, emails, docs) — and then ask the reconstruction to evaluate your draft as that person.

The source packages three variants: "ask the board" (clone 4–5 thought leaders in your vertical from their public content — his example roster is Alex Hormozi, Mark Cuban, Andrej Karpathy — as a strategic sounding board); "internal focus group" (have the model interview to clone the end user of what you are building, for anything customer-facing); and "clone your manager" (for day-to-day work, reconstruct the person who will actually approve it). He reports an Anthropic growth lead cloned his manager from her blog posts, Slack, and emails to get her feedback before submitting.

What distinguishes this from its neighbours is two axes at once. Unlike the Adversarial Planning Council, the personas are specific named real people reconstructed for fidelity, not engineered adversaries built for opposition — and the target is your near-finished output, pre-submission, not an idea pre-build. Unlike Advisor Mode (a smarter model consulted when stuck), the clone supplies not more capability but a particular person's taste and priors. The honest caveats are real: a clone is a reconstruction of a judgment, not the judgment — it inherits whatever the corpus reveals and confabulates the rest — and cloning a real individual from their private messages raises consent/provenance questions the source does not address.

Claims


Linked from