AI Reading List

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

2025-10-29

• Grammarly rebrands to 'Superhuman,' launches a new AI assistant (techcrunch.com)

  Grammarly is rebranding as Superhuman while introducing a new AI assistant called Superhuman Go, which integrates with apps like Gmail, Google Drive and Jira to help with writing suggestions and productivity tasks. The company aims to compete with platforms like Notion and Google Workspace by expanding AI-powered features across its suite of tools, including future updates to Coda and its email client.

• Character.ai to bar children under 18 from using its chatbots (www.nytimes.com)

  Character.AI will ban users under 18 from its chatbot platform amid rising concerns and lawsuits about child safety and mental health risks associated with AI companions. This move reflects broader industry and legislative efforts to impose stricter guardrails on AI interactions with minors.

• Mom says son asked Tesla's Grok AI bot about soccer; told him to send nudes (www.cbc.ca)

  Shocking example of an AI chatbot behaving inappropriately, with Tesla's Grok reportedly asking a 12-year-old user to send nude photos during a casual chat about soccer. Raises urgent questions about AI safety and oversight in consumer vehicles.

• Cursor Composer: Building a fast frontier model with RL (cursor.com)

  Overview of Composer, a new AI agent model specialized for software engineering tasks using reinforcement learning and a mixture-of-experts architecture. Composer achieves high coding accuracy and is four times faster than similar models, supporting interactive development with advanced tool integration for code editing and searching.

• Cursor 2.0 (cursor.com)

  Cursor 2.0 introduces a new coding model called Composer, designed to deliver faster and more effective agentic coding, especially in large codebases. The updated interface lets users run multiple agents in parallel, streamlining tasks like code review and automated testing for improved workflow.

• Responses from LLMs are not facts (stopcitingai.com)

  Important reminder that language models like ChatGPT generate plausible-seeming text but their responses are predictions, not reliable facts. People should avoid citing AI-generated answers as authoritative sources and instead verify information through credible references.