AI moves into cameras, agents and robotaxis

Friday, September 25, 2026

AI is moving closer to the moment a person makes a decision. Apple is testing camera alerts that describe what is happening instead of only saying motion was detected. Microsoft is putting chat, code, and long-running agents into one Copilot app. Google is giving its enterprise assistant a face that can talk back in real time. Waymo is now offering driverless rides in Denver. These are different products, but the common shift is that AI is becoming a layer around ordinary tools and places. That can make small tasks easier, but it also makes permissions, availability, and mistakes more visible. Start with a low-risk use case. Ask what the system can see before you connect it. Check the result against the original source. Keep a human approval step before anything sends, spends, publishes, or changes a record.

Today's top stories

The Verge

Apple Home cameras can describe what they see

The Verge's review compares Apple's new HomeKit Secure Video features with Ring and Google Nest. Instead of an alert that only says motion detected, an AI-generated description can summarize what the camera sees. Availability depends on compatible hardware, plans, and rollout.

Why it matters Faster context can help someone decide whether an alert deserves attention before opening a video clip. Treat the generated sentence as a shortcut, not proof. Tune alerts around events you care about, then check the clip when the decision matters.

The Verge

Microsoft puts chat, code and agents together

The Verge reports that Microsoft's new Copilot app brings Home, Code, and Autopilot into one interface. Home combines chat and Cowork. Code can help create an app, tracker, dashboard, or automation in a sandbox. Autopilot is designed for recurring tasks and follow-ups. Availability and enterprise permissions vary.

Why it matters A small team may get one front door for asking questions, building a first version, and managing repeat work. Start with a non-sensitive task in a sandbox. Review what the agent can see, what it can change, and whether it keeps running after you leave.

The Verge

Gemini Live Avatar gives an AI a face

The Verge reports that Google's Gemini 3.8 Live can respond with an animated persona, lip-sync, facial expressions, 97 languages, and on-screen information. The report says Live Avatar is currently available only to Gemini Enterprise customers.

Why it matters A face can make a live assistant feel more natural, but it does not make an answer more accurate. Treat presentation as separate from trust, and check the source behind the response before acting on it.

New in Robotics

KUNC
Waymo rides are now available in Denver
KUNC reports that Denver residents can order a driverless Waymo across about 60 square miles, with no Denver International Airport coverage. A reporter's ride braked and changed lanes smoothly, but winter, construction, emergency vehicles, and edge cases remain open questions.

LLM Releases and New Features

The Verge
Live AI is becoming a visual product
Gemini Live Avatar pairs near-real-time conversation with an animated persona and on-screen information. The source reports enterprise-only availability, so do not assume this feature is in a personal Gemini account.

Courses and Learning

Microsoft Learn
Start with Microsoft's beginner AI module
Use AI for everyday tasks is an official Microsoft Learn module marked Beginner with eight units and no prerequisites. It covers generative AI, personal and workplace use, and effective prompts. Start with one task you already understand, then compare the AI output with your own answer.

Events

IEEE/RSJ IROS 2026
IROS 2026 starts in Pittsburgh this weekend
IROS 2026 runs September 27 through October 1 at the David Lawrence Convention Center in Pittsburgh. The official registration page lists September 18 as the regular registration deadline and late registration rates. Check current availability and pricing before making plans.

Back to daily archive Previous issue 2026-09-24