CLAUDE 5.5 IS BRILLIANT, CONFUSING, AND VERY HUNGRY

Summary from the TLDR Dev newsletter

Opus 5.5 leads an independent intelligence index while costing less per token than its predecessor, but Sonnet 5.5 can consume far more output tokens at maximum effort. The practical takeaway is to choose the effort level before choosing the model because it often has a larger effect on speed and total cost, while new cyber safeguards may also redirect or block ordinary work.