Hi @MathieuA 
Glad to see you here, and thanks for your feedback
That’s a great idea, I’ve created a topic:
I’ll run Fable on it at some point.
The dashboard is planned! For the scenes, I agree too, I think it’s a good topic after the dashboards.
Indeed, this demo still exists, but I intentionally removed it from the site to see if it improved the conversion rate.
Today, there are only two calls to action on the homepage:
- Install Gladys
- Buy a starter kit
The idea was to avoid having three different CTAs, which risked diluting the attention of visitors and making the page less effective.
I’ll discuss it with Claude to see what he recommends
I’ve thought about it, but not in this form, and it allows me to put a point back on the philosophy of Gladys 
Home Assistant targets a developer audience, with infinite configurability. It’s very appealing to a dev, but this audience is a niche, already very well served. Gladys targets the next audience, much broader: people who want a clean, unified experience that works right away. And to offer this experience, the core of Gladys must keep control over the UI.
So yes to integration widgets, but in the same way that the configuration pages of external integrations are described via a JSON schema, the integration would describe what to display (values, graphs, buttons…), and it’s Gladys that decides how. No custom HTML, no iframe.
It’s exactly the model of iOS widgets: Apple does not allow any arbitrary code on the home screen, everything is declarative. As a result, all widgets are consistent, fluid, dark mode compatible, and no one finds the ecosystem lacking in richness.
The trap of the « in a first step » with the iframe is that as soon as an escape hatch exists, the ecosystem organizes around it and you can never remove it again without breaking everything. A temporary iframe does not exist 
And last point, it ties in with your idea of an AI-generated dashboard: it’s precisely because the widgets would be validatable JSON that an AI could generate a reliable and safe dashboard.
And if tomorrow a type of widget is missing from the vocabulary, we add it to the core and all integrations benefit from it at once, with a consistent UI.
I’ve created a topic: Permettre aux intégrations de déclarer leurs propres widgets de dashboard (schéma JSON)