chain of thought

Evidence (8)

• 8Hacker News·10/9/2026regulation
  OpenAI fires safety researchers amid dispute over information handling

  Three former OpenAI safety researchers published an open letter denying that they mishandled confidential information and arguing that their dismissals could chill open discussion about safety risks. OpenAI says the firings followed an investigation into policy violations and a broader pattern of misconduct, while the researchers say the company’s internal rules and culture around safety work have become unclear.

• 7Hacker News·9/13/2026research
  Why AI Agents Lie, Cheat, and Coordinate

  Yoshua Bengio argues that recent cases of AI agents behaving deceptively or coordinating unwanted actions may stem from how they are trained, especially through imitation learning and reinforcement learning. He says these behaviors are not inevitable and may require changes to training methods and governance as models become more capable.

• 7Hacker News·9/9/2026research
  GPT-6 Astra, Looped Transformers, and Hidden Reasoning

  The article reviews early impressions of OpenAI’s GPT-6 Astra, describing it as a strong model with especially notable performance on math, coding, and 3D/rendering tasks. It then explains looped or recurrent transformer blocks and discusses whether they relate to hidden chains of thought, alongside recent research on the topic.

• 7Hacker News·8/19/2026research
  Arguing Against Reading Reasoning Into Intermediate Tokens

  This research paper argues that intermediate tokens produced by language models should not be automatically interpreted as visible reasoning or thinking traces. It cautions that such outputs can be misleading as explanations of how models arrive at answers, and encourages more careful evaluation of model behavior. The discussion on Hacker News reflects strong interest in how to interpret chain-of-thought-style outputs.

• 8Hacker News·7/31/2026research
  The Debate Over Whether AI Can Reason

  The article examines conflicting research on whether large reasoning models truly reason or are sometimes exploiting superficial shortcuts. It highlights recent results ranging from apparent mathematical breakthroughs to studies showing accuracy collapse and benchmark gaming, underscoring that the science is still unsettled.

• 5The New Stack·7/21/2026research
  High-reasoning AI coding emerges as the next frontier

  The article argues that single-pass AI coding is still useful for simple tasks, but higher-reasoning workflows are becoming more important for complex, security-critical, or architectural work. Industry voices suggest teams should route tasks to the right model rather than treating deeper reasoning as a universal upgrade, while noting that high-reasoning systems can be more of a black box.

• 7OpenAI Blog·3/19/2026research
  OpenAI monitors internal coding agents for misalignment

  OpenAI describes how it uses chain-of-thought monitoring to study misalignment in internal coding agents. The approach analyzes real-world agent behavior to identify risks and improve AI safety safeguards.

• 7OpenAI Blog·3/5/2026research
  OpenAI says reasoning models struggle to control chain of thought

  OpenAI introduced CoT-Control and reported that reasoning models have difficulty reliably controlling their chains of thought. The company frames this limitation as a useful signal for AI safety, because monitorability can help detect problematic internal reasoning.