thinking

Visualization showcasing Claude's different thinking intensity levels, generated using Opus 5.5 and Sonnet 5.5.

TRACES is a new AI benchmark by Apodex designed to measure discoverative AI by focusing on the problem-solving process rather than memorized answers, challenging traditional benchmarks.

An observation that Claude Opus 5's max thinking leads to performance degradation on ~20-30% of benchmarks compared to xhigh, contrary to the expectation that more test-time compute improves performance.

The article explores the growing trend of relying on AI for thinking and decision-making, using anecdotes and a reference to a Ken Liu short story to question the loss of human autonomy.

Paul Graham argues that in the near future, choosing not to use LLMs for writing will be uncommon but a mark of those who care about thinking clearly, warning that a divide between 'writes' and 'write-nots' will mirror a divide between 'thinks' and 'think-nots'.

A rhetorical question about the potential decline in societal collective intelligence if people rely on chatbots for thinking over several years.

An essay exploring why thinking out loud with another person produces better understanding and insight than solitary reflection, drawing on cognitive science and philosophy.

The author uses personal experience to introduce a tutorial on architect thinking in the AI era, emphasizing that the ability to understand the underlying essence when abstraction leaks is more critical than tool usage, and shares two modes: assembly thinking and object-oriented thinking.

Sam Altman shares his belief that clarity of thinking, speed, and quality of execution are linked, using writing as a tool to clarify thoughts.

The author shares a quantization recipe for Qwen3.6 27B that makes the model use significantly fewer thinking tokens while still producing correct answers, leading to faster inference on math benchmarks.