Dan Shipper of Every says Anthropic's Opus 4.8 beats GPT-5.5 on coding and writing benchmarks · Digg