1 articles with this tag
Uber runs 70% of pull requests via agents and cut cost per session 52% by capping context, tuning cache TTL and routing subagents to cheaper models.
