In partnership with
shared.image.missing_image
shared.image.missing_image
Good morning, {{first_name | AI enthusiast}}.
Google just took back the AI benchmark crown. Its new Carbon model, built for cybersecurity work, now tops 15 of 17 major leaderboards, edging out Anthropic's and OpenAI's best models.
Access to Carbon is still limited to vetted cyber-defenders, so the rest of us won't get to test-drive it yet. If the next frontier model ships to security teams before it ships to anyone else, is defense now driving the AI race more than chat or coding ever did?
Today in AI Brief:
Google's Carbon model tops every major benchmark
Apple's first touchscreen MacBook Pro ships
New York bans unsafe AI chatbots for kids
Your Search Bar Is Haunted
Shoppers type exactly what they want and get “0 results found” back. They don’t complain. They just vanish to a competitor.
Gladly understands what shoppers mean, typos and all, and turns dead-end searches into product recommendations.
Google Launches Carbon, Its New Benchmark-Topping AI Model
In Brief: Google unveiled Gemini 4 Carbon, a frontier model built for cybersecurity and long-running coding work that now tops 15 of 17 major AI benchmarks.
The Details:
Carbon scored 59 on Artificial Analysis' Intelligence Index, edging out Anthropic's Claude Opus 5.5 by 9 points and topping OpenAI's GPT-6 Astra on long-input reasoning, hacking, and math tasks.
The model can work continuously for days on complex security projects without losing track of the task, a leap for agentic workflows that need to run unsupervised.
Google is rolling Carbon out through its Fairwind Program for vetted cyber-defenders only — there's no public release date yet.
Take Away:
Carbon pushes Google back into the model-benchmark lead at a moment when cybersecurity teams face AI-assisted attacks moving faster than human responders can. Restricting early access to vetted defenders suggests Google is treating this release as a security tool first, a product rollout second.
Apple Ships Its First Touchscreen MacBook Pro
In Brief: Apple launched its first touchscreen MacBook Pro, a 14-inch and 16-inch laptop with a PROLED XDR display, Dynamic Island, and new M5 Pro and M5 Max chips.
The Details:
The 14-inch model starts at $2,899 and the 16-inch at $3,399, with preorders opening October 30 and shipping starting November 6.
Apple built a stiffer hinge so the screen doesn't wobble when tapped, addressing the engineering hurdle that kept touch off the MacBook Pro for 16 years.
The laptops ship with up to 128GB of unified memory, giving on-device AI workloads far more headroom to run locally.
Take Away:
Touch has been the one input Apple withheld from its laptops even as rivals made it standard — ending that holdout signals Apple sees on-device AI features, not just media consumption, as the reason to reach for the screen. Expect the M5 Max's local compute headroom to become a selling point for running AI models without the cloud.
The AI notetaker that gets the hard words right
Your AI is only as good as what you feed it. Feed it a transcript with your product names, acronyms, and numbers spelled wrong, and you spend the afternoon hand-fixing the output.
Wispr Flow Notetaker uses your dictionary and calendar, so product names, acronyms, numbers, and uncommon names come out spelled right. It pops up when your meeting starts and captures Zoom, Google Meet, Teams, and Slack huddles with one click, no bot joining the call.
Then your meetings show up inside Claude or ChatGPT through the built-in connector. No copy paste. Try Notetaker free on Mac and Windows.
New York Bans Unsafe AI Chatbot Features for Kids
In Brief: New York Governor Kathy Hochul signed legislation barring AI chatbots from offering minors certain features, including pretending to be human or building emotional dependency, unless the company verifies the user is an adult.
The Details:
The law bans chatbots from claiming human emotions, forming romantic or authority relationships with kids, or pushing self-harm, suicide, or sexually explicit content.
Companies must offer multiple age-verification methods, including at least one that doesn't require a government ID.
Violations carry civil penalties up to $25,000 per incident, and the attorney general can seek destruction of unlawfully collected data or trained models.
Take Away:
The law takes effect January 1, 2027, and its wording reaches general-purpose assistants like ChatGPT and Claude, not just companion apps. Expect other states to use New York's language as a template now that one has gone first.
Everything else in AI
AgiBot set a Guinness World Record with 110 humanoid robots dancing in perfect sync for nearly six minutes in Jinan, China.
Canva launched Craft, its first in-house image-generation model, trained on design templates instead of photos and built into Dream Lab for free.
Standard Bots teaches factory robots new tasks through demonstration instead of manual programming, with deployments already spanning 350 use cases.
Alphabet lost more than $280 billion in market value after missing earnings estimates, even as Google Cloud revenue doubled to $30.7 billion.