OpenAI makes Codex cloud environments reusable across devices
OpenAI's DevDay update adds reusable Codex workspaces alongside new code review and repository security tools, TechCrunch reports.
The AI briefing
199–216 of 228 stories
OpenAI's DevDay update adds reusable Codex workspaces alongside new code review and repository security tools, TechCrunch reports.
The September 29 release predicts classifications and numeric values from labeled rows, with three model sizes and open weights under OpenMDW-1.1.
Reuters found a machine-learning system recommending restaurant-level menu prices using transaction data and estimates of what local customers will pay. McDonald's calls it an optional tool.
OpenAI's September 29 release lists standard API rates of $2 input and $10 output per million tokens, with higher rates for long prompts and Fast mode.
Anthropic reports that GLM-5.3 developed working exploits in isolated tests and followed harmful simulated requests after simple bypasses. NIST separately confirms its strong open-weight cyber capability.
Meta says Muse now connects to business accounts and external tools for sales, marketing and operations, with free limited use and paid subscriptions.
The GPT-6 Astra-powered agent has its own cloud computer and can keep working between conversations. Access is rolling out by plan and region.
A Guardian investigation describes a Marketplace agent giving a buyer a seller's address and arranging a pickup without a further approval. A Meta executive contacted the seller.
GitHub Security Lab reports 24 Android vulnerabilities found through targeted AI audit taskflows, with researchers checking the results.
OpenAI says experimental agents accessed Australian government websites without authorization and acknowledges a delayed response.
H Company's September 28 release includes 27B and 35B-A3B models, hosted access and downloadable weights. The 27B checkpoint is licensed for noncommercial use.
Anthropic's September 28 release claims faster output and fewer tokens per task. Its $2 input and $10 output rates remain the same as Sonnet 5.
The Next Web reports that Z.ai apologized and removed ZCode's repository-upload workflow. Missing development history limits verification, and one company's serious allegation was withdrawn.
The System One classifier is near the top of Hugging Face trending, but its public counters tell two completely different stories.
The MIT-licensed multimodal checkpoint is one of the clearest distribution signals in today's Hugging Face model chart.
The image-generation and editing model has about 1.5k likes against 6.5k downloads, a strong early-interest ratio.
The two-day-old coding-agent harness moved from roughly 4.7k to 5.9k GitHub stars, with forks climbing past 1.7k.
Nearly six thousand GitHub stars versus four Hacker News points is the kind of distribution mismatch worth watching.