Most Claude Code “thinking keywords” that people still rely on were removed in the November 2025 build.
I verified this by checking the current bundle and comparing behavior with earlier releases. The old hierarchy no longer exists.
Below is a brief summary of what changed and why many prompts now return zero additional reasoning.
The old ladder is gone
Guides from spring 2025 still describe four levels of “thinking intensity.” That structure isn’t in the current build. The lower tiers are deleted, and prompt text no longer controls reasoning depth.
The new system
Extended reasoning is now controlled by an internal session mechanism rather than keyword stacking. Most phrases developers still append to prompts are ignored.
Why this matters
If your workflow depends on the old hierarchy, your prompts won’t behave as expected. This explains inconsistent reasoning depth in recent builds.
Full breakdown with evidence:
https://medium.com/gitconnected/what-still-works-in-claude-code-nov-2025-ultrathink-tab-and-plan-mode-2ade26f7f45c?sk=a9997a7e950d08916128c56649a890ba
Top comments (0)