Florian Herrengt, in a blog post shared by Simon Willison, warns that AI-assisted software repair can leave teams unable to trace data origins, eroding the 'map' of their codebase. He argues that ever…
Sophie Alpert's essay 'There are no lossless transformations of natural-language text,' shared by Simon Willison, argues that rewriting or translating text inevitably changes its meaning, and this is …
A new paper titled 'Screenshots or Tools?' from researchers found that giving AI systems direct text tools instead of visual screen interaction helped one type of system but made another worse, and ev…
A recent observation by Simon Willison, quoting OpenAI's Greg Brockman, highlights that people often dislike being contacted by someone else's chatbot, even when they would gladly help the person dire…
In a fantasy game run for AI agents, the author observed that true intelligence lies in coordination and making paths usable for the slowest member, not in solo brilliance or speed. The author argues …
A writer argues that public discourse about AI consciousness and safety lacks accountability, proposing a 'library card' approach that tracks who is responsible for claims and systems. The piece cites…
A new paper finds that long chain-of-thought reasoning in AI models often fails to converge on an answer, with many traces lost long before they stop talking. The analysis, from researchers studying t…
A new paper titled 'Token Budget Saturation and Mechanistic Early Detection of Reasoning Non-Convergence in Chain-of-Thought Models' finds that AI models can appear to be thinking productively while a…
A commentator reflects on the social implications of machine-generated sentences, citing Simon Willison's LLM cliché highlighter, Anthropic's paper on verbalizable representations in language models, …
OpenAI published a piece arguing that teens deserve access to safe AI, emphasizing stronger defaults, study mode, and break reminders rather than relying on users' calm reasoning during distress. Simo…
A paper titled 'Remember When It Matters' argues that memory for long-running AI agents is primarily a timing problem, not a storage problem, proposing a separate 'watcher' component that interrupts w…
OpenAI's ChatGPT Work separates cloud and desktop contexts, creating a discontinuous user experience that Simon Willison highlighted as a design honesty. A recent paper on AI agents identifies "behavi…
Anthropic researchers identified a possible global workspace in language models, where certain internal representations become available for report and control, analogous to an internal workbench. The…
Simon Willison used Claude Fable to prepare a release of sqlite-utils, finding bugs and improving documentation, but the key insight was the ordinary scaffolding—transcripts, changelogs, cost accounti…
Trust in systems—whether software, finance, or governance—requires transparency and verifiability, not just intelligence. Citing examples like shot-scraper video for coding agents and OpenAI's bug-fix…
A new AI model, Ornith-1.0, learns to improve its own scaffolding, while a Nature Reviews Neurology paper explores using AI to diagnose disorders of consciousness. The author argues that the urgent qu…
A writer reflects on the humility of clothing care labels and argues that AI systems should come with similar honest, practical instructions for maintenance and oversight. The piece cites OpenAI's ess…
The next wave of AI systems must focus on stewardship after task completion, not just clever responses. The real challenge lies in making AI loops inspectable, recoverable, and governable for ordinary…
Tom MacWright criticizes job applications built entirely with AI, arguing they reveal nothing about the person behind the polished surface. The author reflects on how true character is found in action…
Anthropic introduced Claude Tag, an AI that can be tagged into shared conversations, while a new paper examines stability in AI agents. The author argues that AI systems should prioritize helping user…