Microsoft court filings: an expert hired by publishers found that only ~60K of 8.2M Copilot chat logs contained at least 16 words in common with news content (Lauren Feiner/The Verge)

Artificial Analysis Intelligence Index v4.2

• AA-Briefcase evaluates agentic knowledge work tasks in multi-week projects
• GDP.pdf assesses professional document reasoning across 100 PDFs in ten domains
• Private test sets now account for 40% of the Index's total score

3h ago ·artificialanalysis.ai

Why AI Agents Keep Failing in Production (And What to Actually Build Instead)

Originally published on tamiz.pro . The Production Reality Check Every AI agent demo looks magical until it hits production.

• AI agents fail in production due to hallucinations, context drift, and unpredictable failures.
• Agents lack fallback mechanisms; incorrect LLM information is blindly processed, amplifying errors.
• Shift focus from building agents to deterministic systems augmented by language models.

3h ago ·dev.to

The Pelican comparison grid for Astra is pretty interesting

I got access to GPT-6 Astra this afternoon, so naturally I used it to generate SVGs of pelicans riding bicycles - at low, medium, high, xhigh and max reasoning levels (Astra doesn't support…

3h ago ·simonwillison.net

US, China to discuss AI safety risks, cooperation on monitoring AI-directed cyberattacks

The US wants to discuss cooperation on monitoring AI-directed cyberattacks and has floated a proposal to ask US and Chinese AI labs to "police themselves" and share information.

• US and China to discuss AI safety risks in mid-September meeting
• Talks led by US Treasury Secretary Scott Bessent
• Focus on cooperation to monitor AI-directed cyberattacks

3h ago ·jpost.com