Matthew Berman reviews Anthropic’s Opus 4.8 release, including the stated standard price and faster mode. He compares reported coding and reasoning benchmarks while questioning how well individual benchmarks reflect practical experience.
He discusses the research preview of dynamic workflows, in which a coding agent can divide work among parallel agents and check their resultsA multi-agent system contains multiple AI agents that interact, coordinate, divide work, or influence one another while pursuing tasks.. He considers possible uses in migrations, bug hunts and independent review.
Berman highlights the potential for higher token consumption and asks for clearer evidence of how the new workflow compares with a single model. His comments on providers’ capacity and business strategy are presented as speculation.
Watch on YouTube




