Blog

Notes, ideas, and technical write-ups.

Do Larger Reasoning Models Actually Use Tokens More Efficiently?
Starting from an interview claim: a systematic test with open-weight reasoning models.
July 17, 2026 · Research Notes