Article archive
Browse individual source reports, including coverage still awaiting the summary or community analysis needed for our story feed.
5,357 articles · page 108 of 108
Browse topics
Claude's new constitution
10 min read
Anthropic has published a constitution for its AI model, Claude, outlining its values and intended behavior. This document plays a vital role in the model training process and influences Claude’s outputs, which may not consistently align with the constitution's ideals.
Nvidia CEO says AI needs more investment in defiance of bubble fears
1 min read
Nvidia CEO Jensen Huang stated that artificial intelligence is delivering economic benefits across various sectors and requires increased investment, dismissing concerns about a potential AI bubble. He described AI as a five-layer structure comprising energy, chips, cloud infrastructure, models, and applications.
AI is scoring college essays and conducting interviews, a new layer in admissions stress
9 min read
Universities, including Caltech, are implementing AI tools to evaluate college applicants by scoring essays and conducting interviews. While these technologies aim to enhance efficiency and minimize errors in the admissions process, they also raise ethical concerns regarding bias, transparency, and fairness.
The Agentic AI Handbook: Production-Ready Patterns
19 min read
The Agentic AI Handbook compiles 113 production-ready patterns for building AI agents, including strategies like Plan-Then-Execute and Swarm Migration. A significant shift in AI development metrics occurred during the 2025 winter holidays, indicating a transformative change in the field.
Anthropic's original take home assignment open sourced
1 min read
Anthropic has released a version of its original performance take-home challenge, allowing users to compete against Claude Opus 4.5 with unlimited time. Claude Opus 4.5 achieved a performance of 1790 cycles, while Claude Opus 4 reached 1579 cycles, and Claude Opus 4 took 2164 cycles after extensive testing.
Which AI Lies Best? A game theory classic designed by John Nash
3 min read
Gemini 3 utilizes the So Long Sucker game, a benchmark for AI deception, negotiation, and trust, originally designed by John Nash and others in 1950. The game involves four players using colored chips and requires betrayal for a player to win, enabling the assessment of AI capabilities that traditional benchmarks do not evaluate.
Scaling long-running autonomous coding
3 min read
Wilson Lin at Cursor has experimented with a large fleet of autonomous coding agents, coordinating their work to generate over a million lines of code and trillions of tokens. The process involved running planners and sub-planners to create tasks for workers to execute.