All posts
12 posts, newest first.
2026
- Raw tokens fell 39–60%. The bill went up 14–36%.Deep dive
- Harness²: What an Agent Can Learn by Changing Its HarnessDeep dive
- Reachability Is Not Ranking: Finite Beams, Continuous Queries, and One TIGER-Style BaselineDeep dive
- A Semantic Tokenizer Cannot Be Judged Without Its DecoderDeep dive
- Knowing When to Intervene: Six Papers on RL, Search Agents, and RAGPaper digest
- Search Agents Need Better Decision Boundaries, Not Just Bigger ModelsDeep dive
- What a Recommender Can Reach, Remember, and AffordPaper digest
- When Should a Personal Agent Override Your Recommender?Deep dive
- 语义 ID 如何兼顾记忆与泛化,Agent 怎样学会修正?Paper digest
- Semantic IDs, Reasoning Budgets, and Agents That Learn to RevisePaper digest