Christopher Ort
-
Tampered ML: Even the Best AI Models Miss One in Five Planted Research Flaws
A new EMNLP 2026 benchmark pairs 1,858 ML experiment designs with flawed twins. Claude Opus 5 leads at 77.9%, but models flip verdicts when the order…
-
Realistic MCP Faults Halve Agent Quality, and a Bigger Model Doesn’t Help
Injecting MCP tool failures mined from 1,774 server repos cut agents’ task-completion score by 60%. gpt-oss-120B was no more resilient than gpt-oss-20B; the agent framework mattered…
-
Copilot Users Collaborate More, Not Less: Study of 12,000 Employees
Employees who adopted Microsoft 365 Copilot initiated about 1.4 more collaboration hours a month and reached about 2.2 more colleagues than never-adopters, a log study of…
-
Claude Duet: Anthropic’s Persistent AI Agent for Continuous Work
Anthropic has launched Claude Duet, a persistent AI agent built on Claude Opus 5.5 that operates on dedicated cloud virtual machines to manage ongoing tasks like…
-
Harvard Study Finds Claude AI Outperforms Human Peer Review
A Harvard University study reveals that AI-generated reviews powered by Claude outperform paid human experts on key quality dimensions for academic papers. The findings could transform…
-
LLM Medical Prognosis Limits: Why Models Fail Survival Forecasts
Benchmarks show LLMs excel at extracting data from records but struggle with calibrated long-term risk estimates. Discover why prediction drift blocks clinical use and how hybrid…
-
Shadow AI Risks: Securing Developer Endpoints in LLM Workflows
Enterprises secure production AI but overlook Shadow AI risks from unvetted developer tools and personal LLM keys. Discover how AI Bills of Materials and full-lifecycle controls…
-
LLM Reviewers Send 76% of Published Top-Journal Papers Back, Yet Miss Most Planted Flaws
Four LLMs reviewed 200 papers already published in journals like Nature and Cell: 75.6% of reviews demanded major revision or rejection, while planted flaws were caught…
-
Safety Probes Follow Scaling Laws: Monitor Quality Can Be Predicted in Advance
Across 20 open base models and six safety tasks, probe error follows two simple laws. Fitted on small models, they predict larger ones and pick the…
-
Stale CLAUDE.md and AGENTS.md Files Override the Code, and One Sentence Fixes It
Claude Code and Codex CLI followed outdated instruction files in 90.1% of convention choices. One sentence giving existing code precedence raised code-matching choices to 91.4%.