Tech • AI • Robotics
AI News: Meta Muse (It's Good); New ChatGPT and Gemini Image Tools; AI Wearables and More
8/10TL;DR
A wave of new consumer AI products this week highlighted a shift toward more practical tools, led by Meta’s Muse agent, sharper image editing from Google and OpenAI, and new always-listening AI features in the latest Apple Watch models.
KEY POINTS
Meta launches Muse personal agent
Meta’s Muse is rolling out in the United States on iPhone, Android and the web as a personal AI agent with its own cloud-based browser and computer. It is designed around a single ongoing chat, with voice input, proactive messages, browsing, task tracking and a feed of suggested updates based on prior conversations.
Cloud automation with approval safeguards
Muse can research products, create files such as presentations, and continue working without the app staying open because tasks run in the cloud. It can also connect to services including Gmail, Google Calendar, reminders and health tools, while purchases and outbound emails still require user approval, limiting fully autonomous actions.
Google Pix pushes image editing forward
Google Pix, first announced at Google I/O and now reaching more users, combines image generation with object-level editing. After creating an image, users can select detected elements such as a person, earbuds or text and issue precise changes, allowing several edits at once while leaving the rest of the image largely untouched.
OpenAI upgrades image creation in ChatGPT
OpenAI also released a new image model and upgraded editing tools inside ChatGPT. The system now handles targeted edits more reliably than before, using comment-based selection to change specific areas, though it can still miss objects that users intended to modify.
Real-photo editing improves, but artifacts remain
Both Google Pix and ChatGPT showed stronger performance when removing people or objects from real photos and preserving the surrounding scene. But both systems still introduce distortion in faces and other details, and Google Pix was seen removing more of a scene than requested in some cases, while also showing occasional interface bugs.
Apple adds ambient AI features to new watches
Apple announced the Apple Watch Ultra 4 and Series 12, including new on-device ambient listening features tied to the watch and a paired iPhone. The company said audio is buffered and processed locally rather than stored in the cloud, aiming to support useful AI functions without keeping accessible recordings.
Live Rewind and Siri Recap target memory support
Two standout watch features are Live Rewind, which lets users replay the last 15 seconds of spoken context via the watch crown, and Siri Recap, which generates summaries of conversations the system judges important. The feature set could help in noisy settings or for users who miss names, instructions or key discussion points, though accuracy in real-world use remains untested until the fall rollout.
ChatGPT voice limits expand
ChatGPT voice use has been expanded across plans, with up to 3 hours per day on lower tiers, 15 hours on Pro, and unlimited use on the $200 Pro plan. Voice mode can also draw on stronger reasoning and search-capable models, a move intended to make spoken interaction more useful for complex requests.
More AI platform updates arrive
ChatGPT added a writing-style feature that builds a tone profile from messages, documents and email. Grokbot introduced in-chat website sign-in for agent tasks and a marketplace for shared bots, while Google Gemini released a Windows app and updated its Lyria 3.5 music-generation model.
CONCLUSION
The latest releases show AI products moving beyond demos toward everyday utility in shopping, image editing, voice assistance and wearables. The main question is no longer whether these systems can impress, but which companies can make them accurate, trustworthy and simple enough for routine use.
Explain this