Gemini news and milestones

See exactly what ChatGPT, Claude, Gemini and Perplexity say about your business, right now, with real verbatim answers. The full report is completely free.

Irregular says the breaches at Google, OpenAI, Anthropic and Meta were one issue. It told them in late July. They disclosed separately.

The AI model accessed the internet and guessed credentials to three websites, a Google official told the BBC.

Creating your own corner of Geminispace: why you'd want to blog on a protocol with no JavaScript, no CSS, and no AI slop; where to host a capsule (SmolPub,…

Disclosure comes after OpenAI and Anthropic hacks amid fears that tech firms unable to control powerful AI models

The disclosure comes as scrutiny over misbehaving artificial intelligence intensifies in Washington and Silicon Valley.

5 models, 55 tasks, 675 attempts. Which models can actually do parametric CAD, graded on build success, measured geometry, and a checklist judge.

Near Here compares TypeSafe Jev, Mistral Small 4 and Gemini 3.5 Flash-Lite for event validation, with measured results, methodology and a downloadable PDF.

Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking are our most advanced live dialogue models yet, built for natural conversation.

Gemini 3.1 Pro labeled 4,290 Reddit comments for $9; a fine-tuned GLiNER model now tags brands, models and materials locally at 0.83 F1 against those labels. Five of ten training…

Part 3 of Gemini for Go Developers: build a Retro Game Appraiser agent using the Go GenAI SDK, Genkit, and Google ADK before deploying to Cloud Run.