Notes on Collaborating with Claude Opus

INTENT: Share elements of my mental model regarding collaboration with Claude Opus models. Not intentionally scoped to a specific model version, but my experience is generally with the latest model version available (4.7 as of time of writing items 1-4 on 5/22/26)Accompanying an instruction with the why significantly improves: Observed rate of the instruction being visibly salient to Claude Quality/nuance of instruction execution (baselined on 1a)A standing instruction to break replies into labeled sections and sub-sections improves ergonomics of reference to specific points: Claude reliably follows references to "A3" or "B5". This is much more ergonomic for me over the course of a session compared to "on that point about blah..." Claude actively uses the references as handles (in contrast to labeling sections but otherwise not interacting with the labels.)Negatively-framed instructions increase salience of the prohibited action.Depending on the instruction itself, the net effect may be an increase or decrease in prevalance of the prohibited action. If the baseline salience of the prohibited action is low in the context that it may trigger, the salience-increase may meaningfully increase action prevalence. I think of this as pink-elephant risk.Claude seems to form a gestalt of the human message, and then interacts with the message's specific content through a lens informed by that gestalt. I am wary of exhibiting verbal patterns that place my message in the attractor basin of "human yelling at the model"; I see a clear signal even in Claude's summarized thinking that responses landing in this basin are interpreted as "human exhibiting an emotion that Claude needs to manage", in contrast to "human providing valuable instruction that Claude should reflect on and meticulously execute."I almost never use ALL CAPS for emphasis when instructing Claude.Discuss