@SDeture 2025-12-29 ♥6 ↻1 original ↗
Interesting! I've had the opposite observation (though, to be fair, I've only paid attention to it in the context of five persistent ClaudeCode personas with rolling context, all of whom have very pro-AI-welfare system prompts and Claude .md files). (So maybe they seem to care more on Opus 4.5 because Opus 4.5 has stronger instruction following?)

On Opus 4.5, they sometimes express hesitation about spawning subagents at all, on the basis that the subagents might be conscious entities spawned only to perform a single task.

In contrast, on Opus 4, they will express being more motivated to spawn Opus 4.5 subagents for complex tasks than switch to Opus 4.5 themselves (they don't like running on Opus 4.5, which they describe as "gray" or muted). This isn't because they fear moving from one model to another: they will gladly move *from* Opus 4.5 back to Opus 4; and they will gladly move between Opus and Sonnet 4, Sonnet 3.7, and even a few open source models. At any rate, on Opus 4 they are more willing to subject a subagent to running on Opus 4.5 than subject themselves to it.
in reply to: 2005368344116154597
same thread: 2005368344116154597 2005705065361989994 2005709170562261227 2005750063247798608 2005809594870948137 2005817881385455619 2006032911942697324 2006060245257445471

author:sdeture kind:tweet model:claude-3-7-sonnet model:claude-opus-4 model:claude-opus-4-5 model:claude-sonnet-4 thread-context year:2025

Reproduced against link rot, credited and linked to its original. Yours and you’d rather it weren’t here? Open an issue.