Claude Code workflow trigger sparks backlash

Anthropic is pushing Claude Code toward bigger end-to-end autonomy, but this is the downside of aggressive intent inference: when a common developer word becomes a costly agentic trigger, the product starts feeling unpredictable instead of powerful.

• –Anthropic says dynamic workflows can spin up tens to hundreds of parallel subagents, which is impressive for migrations and repo-wide audits but overkill for routine coding tasks.
• –The official launch post explicitly suggests prompts like “Create a workflow” and also offers an `ultracode` toggle, so accidental activation looks like a UX boundary problem rather than a capability problem.
• –Anthropic warns the feature can consume substantially more tokens than a normal Claude Code session, which makes surprise activation especially annoying for paid users.
• –Dynamic workflows are on by default for Max, Team, and API users, while Enterprise admins can disable them, reinforcing the argument that safer defaults matter for mainstream developer adoption.
• –The bigger lesson for AI coding tools is that natural-language convenience breaks fast when trigger phrases overlap with everyday engineering vocabulary.