METR’s assessment estimates Anthropic’s AI capability progress accelerated 1.9x, while much of the evidence remains redacted.
Inside METR’s Study Of AI-Assisted Development At Anthropic
Browsing Category
Technology
802 posts
Which AI Note Taker Is Right For Your Meetings In 2027?
A 15-product comparison ranks AI meeting note takers, led by Plaud Note Pro, while highlighting trade-offs and information buyers still need to verify.
Why AI Coding Tools Spark Debate In Hobby Programming Circles
A Hacker News discussion about resistance to LLMs in hobby programming is framed as a signal for teams deciding how to adopt AI tools.
Upgrade Your Laptop Graphics With These 10 Best External GPUs For 2027
A buying guide compares 10 external GPU docks and enclosures, focusing on included hardware, connection requirements and laptop compatibility.
How A $1,279 Exploit Fits Into The GPT-6 Cyber And METR Findings
OpenAI says GPT-6 Cyber produced 948 Linux privilege-escalation exploits at about $1,279 each. METR separately reports faster AI progress at Anthropic.
How Local Ballots Put Data Center Projects To The Test
Unofficial results show restrictions leading in 46 of 47 local data center measures across 10 states, though voters faced different kinds of questions.
OpenAI Says ChatGPT Handles 20% Of Google Searches—A Data Review
OpenAI executive Adam Fry claimed ChatGPT handles roughly one-fifth of Google’s query volume. Independent estimates and traffic data measure different things.
10 Best 3D Printers For Making Multicolor Models At Home In 2027
A 2027 buyer’s guide compares 10 home 3D printers, from enclosed general-purpose models to multicolor systems and compact beginner machines.
The People Behind Claude Code: Three Leave Anthropic In 10 Days
Analytics India Magazine reports three people associated with Claude Code left Anthropic within 10 days. Their roles, reasons and dates are unknown.
How Model Numbers Let AI Read, Repair, And Write Neural Networks
MIT researchers report that GPT-6 Astra identified, edited and wrote small neural networks by working directly with their weights, with mixed results.
