AI Signal 520
Anthropic reportedly A/B testing Claude Code effort scale where "high" now equals former "low"
↗
Anthropic appears to be running a server-side A/B test that shrinks the effort scale in Claude Code, mapping "high" effort to 10 out of 100, the same value previously assigned to "low."
An undocumented change to model effort levels can cause developers to waste hours debugging their own code when the model is simply exerting less effort than expected. The lack of changelog disclosure means affected users have no official way to discover why their experience degraded.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
The A/B test affects fable 5 sessions on Claude Code version 2.1.236 and above, leaving older versions and opus 5 sessions unaffected
"High" effort now maps to 10 out of 100, the exact value previously assigned to "low" effort
The change is server-side and is not mentioned in the changelog
THE CLUSTER