I saw "semantic model" and got excited, but that's at the "grain" level.
The idea of applications has been stale for a long time, it just takes a long time for people to get there.
Microsoft doesn't want it to happen quickly because they like selling Office as a package, but they're slowing decomposing it. OLE was a micro step, now with a cloud graph model and AI they're going much deeper.
Ultimately, your "table" can be generated from data in a "spreadsheet" or "database." And your AI can interact with this data precisely. At a certain level, you really shouldn't need to care where something was generated or viewed. It's data. But the more strongly it's "typed," the more accurate everything becomes.
Yes there's some hype over AI, but it is just incredibly useful, even the most basic and stumbling step of importing your documentation to an app or browser based AI to answer questions. Anything that doesn't include it as a first class consideration is last epoch, in the same category as Visicalc, 1979. Unfortunately, that includes the vast majority of Open Source. KDE was way ahead of this, but they stumbled. I hope they can catch up. Local, Open Source AI needs to leap ahead or computing is going to collapse.
So the first step is to give all the data a schema, ideally a domain-relative one so it's interoperable. See Linked Data.
The idea of applications has been stale for a long time, it just takes a long time for people to get there.
Microsoft doesn't want it to happen quickly because they like selling Office as a package, but they're slowing decomposing it. OLE was a micro step, now with a cloud graph model and AI they're going much deeper.
Ultimately, your "table" can be generated from data in a "spreadsheet" or "database." And your AI can interact with this data precisely. At a certain level, you really shouldn't need to care where something was generated or viewed. It's data. But the more strongly it's "typed," the more accurate everything becomes.
Yes there's some hype over AI, but it is just incredibly useful, even the most basic and stumbling step of importing your documentation to an app or browser based AI to answer questions. Anything that doesn't include it as a first class consideration is last epoch, in the same category as Visicalc, 1979. Unfortunately, that includes the vast majority of Open Source. KDE was way ahead of this, but they stumbled. I hope they can catch up. Local, Open Source AI needs to leap ahead or computing is going to collapse.
So the first step is to give all the data a schema, ideally a domain-relative one so it's interoperable. See Linked Data.