AI Reading List

Daily curated list of news, articles, and other interesting posts related to AI.

2025-11-02

• Why do AI models use so many em-dashes? (www.seangoedecke.com)

  Interesting take on why AI models frequently use em-dashes, with the argument that the trend likely comes from the digitization of older print books that feature this punctuation far more than modern writing. The author finds structural and dialect-based explanations unconvincing, suggesting that changes in training datasets are the main factor.

• Context engineering (chrisloy.dev)

  Interesting overview of how the practice of "prompt engineering" is evolving into "context engineering" as large language models are used for complex decision-making. The article explains how expanding and structuring the context window with various types of information helps improve reliability and output, but also introduces new challenges like increased complexity and risk of hallucination.

• Tongyi DeepResearch – open-source 30B MoE Model that rivals OpenAI DeepResearch (tongyi-agent.github.io)

  Interesting overview of Tongyi DeepResearch, an open-source AI agent that matches proprietary tools like OpenAI's DeepResearch in complex reasoning and web tasks. The project introduces a fully synthetic, scalable data generation pipeline and formal approaches for training agents, aiming to boost reliability and performance for advanced information retrieval and decision-making.