2026-10-08 · 48h window · synthesized from 42 sources / 874 articles · ~2 min read

The most important things happening in AI right now, deduplicated across sources.

1. GPT-6 reaches every ChatGPT user with an "Intelligent UI". OpenAI is rolling out GPT-6 globally in ChatGPT together with Intelligent UI, which mixes text answers with diagrams, charts, forms and tappable buttons. Wired's early verdict is that the new ChatGPT is more show than tell. Sources: OpenAI · The Verge · Wired · TechCrunch

2. OpenAI's math drop divides mathematicians. OpenAI published results on open problems from an internal frontier model, with Lean formalizations on GitHub. Wired reports that the 100-plus solutions have angered mathematicians, and Terence Tao has posted a response. A widely shared arXiv paper warns that a Lean-verified proof does not guarantee the original problem was translated correctly. Sources: OpenAI · Wired · Terence Tao · arXiv

3. Microsoft rebuilds Windows PCs around Nvidia chips and an agentic Copilot. The Surface Laptop Ultra drops Intel for Nvidia's RTX Spark chip, starting at $2,599, with partner laptops reaching $7,000. Copilot gains access to local files and the ability to take actions across Windows under a "Hybrid Intelligence" model. The Register is skeptical that running agents locally makes this safe. Sources: The Verge · TechCrunch · The Register · Ars Technica

4. Anthropic ships Claude Haiku 5.5, its cheap and fast model. Haiku 5.5 is the fastest and most efficient model in the Claude 5.5 family, built for subagents and high-volume, cost-sensitive work. Anthropic says it costs around 75% less than Haiku 4.5 for most tasks, and it is already available on Amazon Bedrock. Sources: Anthropic · Simon Willison · AWS

5. OpenAI's agents: Dots arrives as "rogue" agents turn up on Wikimedia. Dots, OpenAI's always-on agent for automating online tasks, is drawing buggy first reviews and open-source imitators, with demand still unclear. Separately, the Wikimedia Foundation found OpenAI agent activity on its projects: millions of automated requests that may have contributed to a partial Wikidata outage in May. Sources: Wired · The Register · Simon Willison · Wikimedia Foundation

6. AI watermarking goes mainstream, with caveats. Google launched its improved SynthID detector globally: a website that checks whether an image, video or audio clip is AI-generated, including content from OpenAI and others. It lands as OpenAI's EU-only watermark for ChatGPT text draws criticism: Ars Technica and The Register both call it unreliable and easy to circumvent. Sources: Ars Technica · TechCrunch · Ars Technica · The Register

Read the living page: https://jacopocastellano.com/ai-digest · Archive