The manager readme failed because the document was never the deliverable

The mechanism here is an ordinary proxy problem. Two things could count as the output of a manager readme, and only one of them is countable: whether a document exists, and whether the manager actually worked out what they do to the people around them. Rands is clear which one paid off for him. The task was to reflect on what mattered to him as a leader, and getting it out of his head and onto the page forced him to think the subject through ^[6]. When the practice spread by imitation among the managers on his team, the file was the only visible part, so the file is what got shipped ^[7].

Then there is the reading problem, which depends entirely on direction. Rands's own entries are not policies, they are involuntary reactions. Hand him a task and he has already re-graded you on every prior task, and that grade quietly sets how much he hovers ^[12]. Let the team drift off a plan he never said out loud and the lists with boxes appear, and the boxes must be checked ^[13]. Read sideways by a peer, that is useful context. Read downward by someone whose compensation you set, the same sentence arrives as an instruction with no verb attached. Rands says the recipients were rightly furious and that the document can read like a manifesto ^[9][10]; what they were handed was a list of ways to not irritate their boss, with no standing to change any item on it.

Which is why the feedback pattern is the real finding. Of the three sections he names, habits, values, and Nuance and Errata, exactly one drew substantive response despite all the noise around the format ^[14][15][19]. It came from a General Counsel at a previous company, quoting the line about going quiet in groups larger than eight ^[16], who wrote "Now I understand you" ^[17]. A lateral reader, self-selected, using the text to reinterpret behaviour he had already observed. Nobody with something at stake wrote back.

That fits Rands's own standard, that leaders lead by mandate poorly and by example well, and that people watch how you act and decide for themselves whether it works ^[18]. A readme sent to reports has to compete with the behaviour they can see, and it only ever loses that argument: every gap between the stated principle and the Tuesday meeting is now documented in the manager's own words. Sent to a peer, there is no gap to audit, only a person to understand. The writing survived the fad in good shape. The distribution never had a case.