A 15-question Lexassay test scores 42 AI setups on legal tasks. Claude Opus 5.5 leads; a model built for law scores 8 out of 100.
We Gave 42 AI Setups The Same 15 Legal Questions. The Model Built For Law Scored 8 Out Of 100
Browsing Category
Artificial intelligence
225 posts
How GPT-6 Astra Edited Another AI Model’s Weights To Fix A Loop
A November 2 blog post reports that GPT-6 Astra edited two Qwen3.5-0.8B layers, raising its MATH-500 score while reducing, not eliminating, loops.
The End Of The AI Flat Rate? Musk Says Grok Bot Cost Too Much, OpenAI Tightens The Rules. What Changes For You
Elon Musk said serving Grok Bot cost too much relative to revenue as OpenAI added account-sharing limits and a shared ChatGPT-Codex allowance.
Who Pays For AI’s Data Center Expansion? Trace The Money
Big Tech’s data center spending has surged. Cash, debt, leases and outside investors are helping pay for the AI build-out, with risks spread across sectors.
How Google’s Genie 4 Turns Prompts Into Worlds You Can Explore
Google DeepMind’s Genie 4 turns text prompts into interactive audiovisual scenes, but access is limited and the worlds can drift over time.
Six Ideas From My Book Dots, Bots, And Math! That Explain This Month’s AI News
Thorsten Meyer connects six ideas from his new book to recent AI headlines, while stressing that the book offers a framework, not predictions.
Top Links 1253: AI Crowding, US-China Investment Decoupling, And Russia’s Plague Death And Hyperpolitics
Adam Tooze’s Oct. 13 Chartbook post points readers to links on AI crowding, US-China investment decoupling and Russia’s plague.
AI Can Spot the Story. Can It Close the Deal?
Firmulate’s AI company experiment finds that models can spot crises and resist manipulation, yet still miss the deal. Enterprises can wargame their own business.
AI’s New Management Test: Can a Model Close the Deal?
Moonshot’s Kimi K3 placed second in Firmulate’s live company test, beating three Western models and raising a sharper question than chat quality alone.
The Unexpected Benchmark: Why a Do-Nothing AI Gets 26 Points and What It Means for Business Trust
A new AI benchmark shows even passive models get 26 points, highlighting trust, honesty, and data reading as key to real-world business AI performance.
