Claude Operator: Prompt to Autonomy · 18 min · 130 XP
Comparing, resolving, and synthesising
Use multiple workers to surface assumptions, not to hold a vote.
Asking two specialists to analyse the same decision from named perspectives — "argue for keeping this dependency", "argue for removing it" — is one of the highest-value uses of delegation. Done well, it surfaces the assumptions each position needs, which is the part you couldn't easily produce alone.
Done badly, it becomes a vote. Two agents agreeing is not evidence: they may share a source, a blind spot, or the framing of your question. Read the reasoning, not the tally. What you want from the exercise is a clear view of what would have to be true for each answer to be right.
When findings actually conflict, resolve by tracing both back to primary evidence. One will usually turn out to rest on something checkable: a file that says what it says, a benchmark that ran, a version that is or isn't installed. Decide on that, and record what remains genuinely uncertain rather than declaring the matter closed.
Then synthesise without flattening the disagreement. A summary that reports three workers' findings as one smooth consensus has destroyed the most valuable information in them. Keep three sections: what they agree on, where they conflict and why, and what you've decided given that.
Practice. Ask two specialists to analyse one real choice from different named perspectives and compare what assumptions each relies on. Take two conflicting claims and trace both to primary evidence, citing what decided it and what's still uncertain. Then combine three worker reports into one synthesis with agreement, disagreement and decision sections, preserving confidence levels.
Loading your workspace…