Why the World’s Most Important AI Dataset Is Probably Wrong
The dataset that launched modern deep learning was assembled in 2006 and encodes assumptions that are now embedded throughout AI. Here's what…
Mira Okonkwo
11 stories
Inside the Company Building AI That Explains Itself
Cohere has been building AI that explains its reasoning while everyone else was racing for benchmarks. Their enterprise bet is starting to…
AI Is Getting Better at Knowing What It Doesn’t Know
Frontier AI models are getting better at expressing uncertainty when they should. The calibration problem hasn't been solved — but it's improving…
The Model That Refused a Government Request — and What Happened Next
A national security agency requested model weights from a major AI lab through formal legal process. The lab said no. What happened…
The Open-Source AI Safety Lab Nobody Is Talking About
EleutherAI has quietly built the infrastructure for open AI research — the datasets, evals, and interpretability work that the whole field depends…
Inside the Six-Month AI Governance Crisis Nobody Reported On
Three major AI labs quietly modified deployments after internal safety evaluations flagged unexpected capability jumps. This is what happened.
Anthropic’s Constitutional AI Is Now a Template Other Labs Are Copying
Constitutional AI has gone from proprietary research to industry standard faster than almost anyone expected.
OpenAI’s GPT-5 Has a Reasoning Mode That Pauses to Ask You Questions
Rather than proceeding with ambiguous instructions, GPT-5's new clarification mode surfaces uncertainty before it compounds.
Why Your Company’s AI Policy Is Already Out of Date
Enterprise AI governance documents written in 2023 were designed for AI assistants. They are almost universally unprepared for agentic systems.
The Year AI Stopped Assisting and Started Deciding
Autonomous agents are no longer lab experiments — they're signing contracts, managing infrastructure, and making calls that used to require a C-suite.