OpenAI Dots, Meta Muse, Gemini Live, and Shopify Canvas are moving Artificial Intelligence (AI) from the chat window into software, cameras, online stores, and connected devices. This shift affects more than people who generate text: it also matters to shoppers, people with impaired vision, small businesses, and smart-home users. The products also show how varied an AI agent — software that carries out tasks with some autonomy — can be.
From conversation to action
OpenAI’s Dots is the clearest move toward a general-purpose work agent. You talk to your Dot in one window while watching it operate a virtual machine in another; a virtual machine is a software-based computer kept separate from your physical device. According to a hands-on test of Dots, the agent can use applications such as the Blender 3D tool and GIMP image editor, access your own computer when permitted, and discuss work with you over a call.
That makes Dots look more like workplace software than an occasional shopping assistant. A Dot could, for example, work on a task in creative software while showing you its progress. The same agent can also order dinner, blurring the line between digital coworker and everyday helper. OpenAI plans to support multiple Dots, but the current version provides one, according to the report.
Access is the first substantial limitation. Dots initially requires one of OpenAI’s highest subscription tiers, including the $100-per-month Pro plan. Meta Muse takes a free-access approach instead: according to a comparison of Muse and Dots, anyone with a Meta account can use Muse and gets a dedicated virtual machine. The reports do not establish whether those free terms will remain in place over the long run.
Four routes into daily life
Meta does not want to confine Muse to a browser or a single app. Its open-source Muse Gadgets project provides firmware — the basic software that runs a device — and a software development kit (SDK), a set of tools for connecting software. It is primarily intended for technically experienced hobbyists who want to connect Muse to hardware such as a color E Ink display, a TV-mounted HDMI stick, or a small Raspberry Pi computer.
Meta has also produced 5,000 units of a USB-C device called Home Link and says it will give them to Muse subscribers at no charge while supplies last. Home Link connects Muse to a home network and uses community-built skills to perform tasks such as turning on a light, controlling a television, or sending a document to a printer. Meta explicitly says that people build custom devices at their own risk. The report on Muse Gadgets and Home Link therefore illustrates that openness does not automatically equal ease of use.
Google is taking a different route with Guided Vision in Gemini Live. On compatible Android devices, the feature provides spoken descriptions of whatever you point the camera toward. It can read small print, locate or identify objects, and describe specific details. One concrete daily example is finding an item and then asking Gemini to read its expiration date. Audio cues can help you align the camera if the relevant object is outside the frame.
Guided Vision is particularly aimed at people who are blind or have low vision. It is also available through Google’s TalkBack accessibility service, and the report says users can configure an accessibility shortcut on devices running Android 9 or later. Google warns that the feature should not be used for navigation, safe-travel guidance, or obstacle detection. It is not a replacement for a cane or another mobility aid.
In Switzerland, the move from apps to wearable devices is already visible. Sunrise has added several versions of Meta AI Glasses to its range and plans to provide demonstration units in selected stores, although its announcement says this could take several weeks. Listed models include the Meta Adventurer starting at 269 Swiss francs and the Meta Fury at 269 francs. The Swiss launch through Sunrise also carries a familiar unresolved issue: glasses capable of recording video in public remain controversial.
Shopping and selling through conversation
For shoppers, OpenAI is offering visual assistance rather than a completely autonomous purchase. ChatGPT’s virtual try-on lets you upload a selfie or full-body photo and generate a view of how an item of clothing or an accessory might look on you. You can also upload an image of a product, such as a screenshot from the web. A Favorites feature saves discovered products and generated try-on images in a library.
As a practical example, you can describe a style you want to explore and ask ChatGPT to find the individual pieces. You can also upload a photo of an outfit and request similar items that are available for purchase. OpenAI says the shopping tools are launching globally and use its ChatGPT Images 2.5 model. The claims about more natural lighting, richer textures, and lower generation latency come from the provider and have not been independently verified in the supplied reports.
The product’s history calls for some restraint. OpenAI previously had to change direction after an instant-checkout feature did not perform well. Recommendations pushed proactively by another AI shopping provider were also perceived by some users as closer to advertising than useful assistance. In any event, a convincing generated picture cannot confirm that the real garment’s size, material, or cut will suit you.
Shopify Canvas works on the seller’s side of the transaction. Merchants describe the store they want to the Sidekick AI agent in a chat, while Canvas displays changes in real time. According to Shopify, the result is not a static mockup but a rendering of the actual store. This allows merchants to test interactions, animations, and different screen sizes, while retaining the option to edit individual elements directly.
A concrete professional use case is a small retailer that wants to redesign its online store without editing code for every change or immediately hiring a developer. According to the provider, Sidekick can work directly with the theme files and takes screenshots so the agent sees the same current version as the merchant. Canvas brings earlier Sidekick abilities together, including code, app, and theme changes, in a single overview.
Pros and Cons of everyday and shopping agents
Pros:
- Concrete actions – Dots and Muse go beyond answering questions by operating software and connected devices.
- Improved accessibility – Guided Vision can read small print, describe objects, and answer follow-up questions about a camera view.
- More visual shopping – Virtual try-on helps you assess products before saving them or making a purchase.
- Lower entry barrier – Canvas lets merchants modify a store through conversation and inspect the results immediately.
Cons:
- High or uncertain costs – Dots initially requires a plan starting at $100 per month, while the long-term terms of free alternatives remain unknown.
- Errors with real consequences – A wrong purchase, device action, or visual description matters more than an inaccurate chat response.
- Privacy and observation – Photos, camera feeds, store data, and smart-glasses recordings can contain sensitive information; the sources do not provide a complete Swiss privacy assessment.
- Limited reliability – Google explicitly rules out safety-critical navigation, while Meta says custom Muse devices are used at the builder’s own risk.
What this means for you
For beginners: Your most sensible first step is a limited task whose result you can easily check. You could ask Guided Vision to read small print or use ChatGPT to visualize an item of clothing without triggering a purchase. Verify the expiration date or product information yourself afterward, and avoid uploading images that reveal unnecessary personal details.
For advanced users: You will get more value by placing the agent inside a clearly bounded workflow and reviewing intermediate results. With Canvas, that means testing store pages at multiple screen sizes and checking interactive elements yourself. With Dots, you should monitor which applications the agent uses and decide how much access to your computer the task actually requires.
Despite its approachable characters, Muse Gadgets is currently better suited to people with technical experience. The two reports about Meta’s hardware discuss development boards, firmware, and an SDK, which is not a typical consumer-device setup. A finished product such as Home Link or commercially available smart glasses will be more accessible for most people, but the sources leave availability, supported languages, and detailed privacy conditions partly unresolved.
Together, the four approaches mark a genuine change in function: AI no longer merely describes what you could do but can act through software, cameras, products, and devices. The practical value is already visible in narrow, verifiable tasks, while pricing, availability, and reliability vary considerably. The unresolved risk is that a convenient agent may receive more data and authority than the task actually requires.
Sources
- OpenAI’s Dot agent is enterprise software that can also order your dinner – The Verge, 2026-10-02
- Meta wants your next gadget to be Muse-infused – TechCrunch, 2026-10-03
- Meta open sources code to let you make Muse AI gadgets – The Verge, 2026-10-02
- OpenAI’s new agent is a shot at Meta — but can it compete with free? – The Verge, 2026-10-01
- Google’s new Guided Vision feature can help you read the fine print – The Verge, 2026-10-01
- ChatGPT can now virtually try on clothes for you – TechCrunch, 2026-10-01
- Shopify debuts Canvas, a way to build online stores by chatting with AI – TechCrunch, 2026-10-01
- Sunrise steigt ins Geschäft mit Smartbrillen ein – Netzwoche, 2026-10-02
Image: Helena Lopes via Pexels