Claude Code's binary reveals silent A/B tests on core features

A paying Claude Code user decompiled the binary and discovered a GrowthBook-managed A/B test called 'tengu_pewter_ledger' that silently alters how plan mode generates plans, with variants ranging from full context to a hard-capped 40-line terse format. The author argues that running undisclosed experiments on a $200/month professional tool undermines transparency and responsible AI deployment, as users have no opt-in, notification, or configuration control over these changes.