Artificial Analysis tested 23 models on FrontierCode, a benchmark that grades whether AI-generated repository changes meet maintainer-style review criteria.
FrontierCode: The Path From AI-Generated Code To A Pull Request
Browsing Category
AI & Tooling
617 posts
How To Evaluate AI Models For A News Pipeline: Lessons From 71 Tests
A 71-configuration test found GPT-6 Astra scored highest, while lower-cost GPT-6 Luna was selected for production; two unreliable fallbacks were removed.
Muse Wonder Full Steps Into AI’s Frontier As Muse Wonder Charges
Artificial Analysis scores Muse Wonder Full at 56, while the API model costs $4.52 per benchmark task—far more than some similarly scoring rivals.
AI At Work: Understanding The Dots, Bots, And Math Behind The Shift
A commentary links persistent AI agents, autonomous software and a proposed math result to possible changes in how organizations structure work.
What Mathematicians Should Know About The Lean Theorem Prover: Reliability & AI
Mathematician Thomas Hales surveys Lean’s reliability and AI-driven formalization, while key questions about checking and verification remain.
The Role Of Influencer Scoring In Ecommerce Product Launches
A proposed workflow would rank launch influencers using audience fit, engagement and sales data, then test predictions against results from 10 launches.
OpenAI Is Training Agents In Your Software. Ironclad’s Terms Matter
OpenAI says GPT-6 Astra met 55% of rubric criteria across 11 Ironclad tasks; its time estimates are simulated, and human review remains necessary.
What To Know About Grok’s Lawsuit Against Minnesota
A Baptist News Global headline reports xAI is suing Minnesota over restrictions tied to Grok’s sexualized images, but key case details are unavailable.
A First Look At Claude Haiku 5.5 On AWS
An AWS page titled “Introducing Claude Haiku 5.5 on AWS” names the model, but the available material does not confirm launch timing, access or pricing.
AI’s Productivity Tradeoff: More Work To Review
AI is accelerating output in math, software and legal work, while available evidence points to review capacity lagging behind.
