A system prompt gives a model standing instructions before you type a request. Product coach Paweł Huryn reports that Claude Code’s instructions shrank, and says he added rules to address repeated workflow problems.
Paweł Huryn, in a post on his newsletter The Product Compass, says he measured the difference directly:
“Anthropic removed 80% of Claude Code’s system prompt and was proud of it. I measured the cut: 2,686 words down to 514.”
Who Huryn is
Huryn is the founder of The Product Compass, a product management newsletter. He describes himself as an ex-Chief Product Officer with 15 years in product roles, and now writes and teaches about how AI changes product work.
What he gets right, and where it’s incomplete
Huryn reports a reduction from 2,686 words to 514. His examples describe two different problems: unnecessary permission pauses and action beyond the scope of an evaluative question.
He connects those problems to a training emphasis on long jobs. Reinforcement learning rewards a model for outcomes during training. He summarizes his interpretation as “it was trained to work without you, so it stopped working with you.”
That is Huryn’s own theory connecting a measured prompt change to an observed behavior shift. The causal link is his interpretation, rather than a confirmed explanation from Anthropic. He also cites reactions he says he has seen on his social media timeline, including calls to “bring back 4.6,” as evidence of wider frustration. Those are secondhand and anecdotal, not a survey or Anthropic’s own data, so they show that some users are unhappy without establishing how common that reaction is.
Huryn also cites an Anthropic evaluation table from July 24, 2026, saying Opus 5 scored about three times the next-best model on ARC-AGI-3, a benchmark for solving problems the model hasn’t seen before. That figure comes from Anthropic’s own published results, not Huryn’s independent testing.
What he adds beyond the complaint is a fix. Rather than switching models, he wrote explicit rules back into his own custom instructions to restore the behavior he wanted.
Two public rules address those different failures. “Act. Don’t Ask” tells the model to complete cheap, reversible tasks within the given scope. “A Question Is a Question” tells it to answer an evaluative question such as “how could we improve this?” before implementing changes.
Why it’s notable
Anthropic scheduled auto mode as Claude Code’s default on Pro, Max, and Team from August 14. It uses an automated classifier for routine approvals.
Permission controls and custom instructions solve different problems. Auto mode reviews whether an action is allowed; Huryn’s rules clarify what work he wants done.
What it means for builders
Anthropic’s default instructions are tuned for Anthropic’s idea of a typical user, not for your workflow. If Opus 5 (or any model update) suddenly acts more independently than you expect, don’t assume you have to accept the new behavior or switch models. Write down the specific failure, the exact way it went wrong, and turn it into an explicit instruction in your own system prompt or custom instructions.
Then check whether that instruction changes the next result.
End of article