We rebuilt our harness around decisions and saw a 30.7% increase in token efficiency.
The new harness running Sonnet 5 beat our old harness with Opus 5 using fewer tokens.
Claude opus 5 stopped building websites and started making design decisions nobody asked it to make
Give it a rough brief and it doesn't just fill in sections, it pushes back on your own instincts. suggests a different type scale because the one you asked for clashes with the
Do you want Fable 5.1 to run gpt 5.6 Sol agents?
Or Astra to run pi, opencode, opus etc? Try following with herdrdev: [PROMPT] I want you to act as planner and orchestrator, to do <task_description> where you delegate backend work to gpt sol 5.6 subagents <or insert other
Nex-N2.5 arrives open source: 0.1 behind Opus 5 on one benchmark, less than half on computer use · NexEcosystem
You have reached the end of the archive
All of Claude Opus 5