Claude appears to have stopped showing full summarized thinking traces to users, according to a viral post by researcher @emollick on X. The change eliminates the detailed reasoning summaries that previously accompanied Claude's responses, a feature that provided unique interpretability benefits.
Key facts
- Posted by researcher Ethan Mollick on X.
- Before-and-after comparison shows traces missing.
- Claude's thinking traces were unique among major models.
- Anthropic has not confirmed the change officially.
- Users report inconsistent behavior across queries.
Researcher Ethan Mollick posted a before-and-after comparison on X showing that Claude's thinking traces — the summarized chain-of-thought explanations that appeared alongside answers — are now truncated or missing entirely. According to @emollick the shift represents 'a big loss, both for interpretability (seeing even a summarized thinking trace helps you diagnose errors in a way that you can't otherwise) and because they were insightful.'
These traces were a rare window into the model's reasoning process — a feature most competitors do not expose. GPT-4o and Gemini offer no equivalent, and even OpenAI's o1 reasoning model only shows a brief, abstracted summary. Claude's traces allowed users to spot hallucinations, biases, or flawed logic mid-conversation, making them a critical debugging tool for developers and researchers.
Anthropic has not issued an official statement or changelog confirming the removal. The company did not respond to requests for comment. It is unclear whether this is a deliberate product change, a temporary bug, or an A/B test. Users on Hacker News and Reddit have reported inconsistent behavior, with some still seeing traces on certain queries.
If the change is permanent, it marks a retreat from the transparency that differentiated Claude from GPT-4o and Gemini. It also raises questions about Anthropic's priorities: is the company trading interpretability for speed, cost, or safety? The timing is notable — Anthropic recently closed a $5B Series E and is under pressure to scale Claude's enterprise adoption, where concise answers may be valued over raw reasoning.
Key Takeaways
- Claude may have removed full thinking traces per user reports.
- Loss for interpretability and debugging.
- Anthropic has not confirmed.
What to watch
Watch for an official Anthropic changelog or blog post addressing the thinking trace behavior. Also monitor user reports on X and Reddit over the next week to determine if the change is permanent or a transient A/B test. If confirmed, expect pushback from the interpretability research community.
