Unbiased AI-powered news
A report from OpenAI, Columbia, Duke and the University of Pennsylvania shows increased use of the agentic coding platform Codex. Token data indicate higher reliance among employees, organizations and individuals since early 2026.
forbes.comUse of OpenAI's Codex agentic coding and work platform has accelerated since the start of 2026, according to a report from OpenAI, Columbia, Duke and the University of Pennsylvania shared first with Axios. Among active ChatGPT and Codex users at organizations outside OpenAI, the share using Codex rose from just above 0 percent in August 2025 to approximately 17 percent.
OpenAI employees produced 99.8 percent of their output tokens with Codex, while organizations reached 63 percent and individual users reached 16.5 percent.
Seventy point two percent of Codex users made at least one request estimated to save more than an hour of human work. Twenty five point six percent delegated work estimated to take more than eight hours for a human to complete. The thresholds are model-estimated and based on a 0.1 percent random sample of individual users who opted to allow queries for training.
Non-developers form the fastest-growing user group, though software work remains the core use case. The shift began in earnest at the beginning of 2026, when users began allowing Codex, OpenClaw and Anthropic's Claude Code to interact with desktops, manage calendars, read and write files, control web browsers and execute scripts.
Fewer than 1 percent of active consumers on Go, Free, Pro and Plus plans used Codex in the last 28 days, Axios reported.
Single source — no framing comparison available.
Anthropic sent a June 10 letter to U.S. Senators Tim Scott and Elizabeth Warren stating that operators linked to Alibaba created thousands of fraudulent accounts and ran nearly 29 million exchanges with its Claude model. The letter described the activity as distillation attacks a…
theregister.comThe European Union held talks with the Trump administration about access to Anthropic PBC's most advanced artificial intelligence models. The discussions followed U.S. restrictions on foreign nationals accessing the technology earlier this month.
TechCrunchOpenAI introduced its first custom inference processor, developed with Broadcom and named Jalapeño. The chip targets lower operating costs for running AI models and is still undergoing testing.