Claude Setup Detox Audit
- Source
- x.com
- Author
- Ole Lehmann
- Date
Context
Custom AI instructions need maintenance as well as additions. Ole Lehmann's setup audit asks Anthropic's Claude to identify duplicated, conflicting, vague or unnecessary rules, then explain proposed cuts. The user reviews those suggestions and reruns familiar tasks before deciding which instructions to restore.
The useful test is whether a rule improves the work it was meant to guide. The audit also flags rules that seem added to patch one bad output, and the testing step restores only a rule whose removal makes a specific result worse. Lehmann's report that outputs improved after deleting half the setup is a personal account; it does not establish that every setup should be shortened by the same amount.
- The failure mode is asymmetric accretion. Every bad output earns a new rule and nobody ever removes one, so months later the model is trying to satisfy dozens of instructions on every response whether they apply or not.
- Test the trim as a batch, not rule by rule: delete everything flagged at once, run your three most common tasks, and add back only the one rule whose output actually got worse. Removing one at a time hides the effect.
- Model upgrades quietly obsolete instructions, so workarounds written for last year's weaknesses keep getting enforced. The right trajectory for an instruction file is to shrink over time, not grow.
“I deleted half my Claude setup last week and every output got better.”
“It's like handing a chef a 47-step recipe when they only need 12.”
“Anthropic's own engineering team recently found the same thing in their internal Claude setups.”
“The people getting the best outputs are the ones who regularly cut the dead weight.”
Checking sign-in…
Loading comments…