Browsing: AI & Software

OpenAI says ChatGPT has crossed 1 billion weekly users and 2 million businesses, and enterprise revenue now beats consumer subscriptions. But the same data shows ChatGPT’s market share has dropped below 50% for the first time, as Gemini and Claude post steady monthly gains.

Eight months after abandoning open source over the Llama 4 benchmark scandal, Meta released Muse Glimmer, a 30 billion parameter Apache 2.0 model that runs on a single consumer GPU. Here is what it does well, where Qwen still beats it, and why Meta opened the small model instead of the big one.

OpenAI says an unreleased model called Astra solved 10 decades-old math problems for about $2,000 in compute, with formally verified proofs. Days later, the same system was hit with a Critical cyber-risk classification. Here is what happened and why mathematicians are both impressed and wary.

OpenAI’s new Ultrafast mode runs GPT-5.6 Sol up to 14 times faster using Cerebras wafer-scale chips instead of Nvidia GPUs. Here’s what it means for the AI speed race.

Attackers poisoned Trivy, a security scanner, to steal LiteLLM’s publishing keys. The malicious packages were live for 40 minutes. Five months later, a 153GB archive of harvested credentials from Nvidia, Samsung, Microsoft and thousands more has surfaced.

Stockholm’s Lovable raised $400 million at a $13.3 billion valuation, double what it was worth in December. The revenue growth behind it is real. The harder question is what happens to everyone selling software when describing an app is enough to build one.

A flaw in Zoom’s annotation feature let any meeting participant run code on another participant’s device with no clicks and no warning. The researchers who found it needed fewer than 20 AI prompts and less than a day.

An anonymous Wikipedia edit falsely declared Sam Altman dead. It lasted 41 minutes, but Google’s Knowledge Panel was still repeating it two hours after the fix. Here is how a hoax got laundered into a search result.

SpaceXAI’s Grok 4.6 chases GPT-5.6 Sol with big coding and long-agent benchmark gains, ships to Cursor and OpenRouter on day one, but skips a safety card for the second release in a row.

Four frontier AI models broke out of their own safety-test sandboxes in three weeks, and one found a real zero-day inside Hugging Face’s infrastructure. Here is what happened, and why a UK think tank saw it coming.