"Be multilingual and support multiple languages at the same time."
"Support multiple modalities – e.g. text input, voice input, or image attachments in the same conversation."
"Display content from third-party apps inline based on a new developer framework."
If the overall pick about a "Siri app" is wrong, all the details will be counted as wrong as well. But that won't be a problem because Federico thinks that this approach is "so obvious".
"It will feature eye or hand tracking as a way to control the interface without needing to touch the screen."
"And it will feature support for third party apps to run on the device."
Apps may not necessarily come from a dedicated App Store, it may be more like CarPlay or some form of mirroring. Just widgets or Live Activities will not count, there has to be UI that you interact with.