This will require someone to train a custom LLM built on the OH docs and some working code. It will work better for text based code over Blockly I suspect for no other reason than the chatbots tend to be text focused.
If the videos @hmerk linked to don’t help, open a new thread and I’ll gladly answer any questions and try to clear up any confusion you might have. But at a high level, it’s code like any other code. Instead of writing it out in text though, you pick “blocks” and plug them together. All the cocepts you already know from Rules DSL still apply.
And if Blockly isn’t your thing that’s good too. You can continue to use Rules DSL (not my recommendation) or use JS or jRuby, even in the UI.
This one might be challenging but the rest should be doable I think. But if you want sematic stuff I would think you’d want to use the Model instead of Items page.
I bet if someone published it to the marketplace lots of people would use it. 
And 5.0 too. I listed is above. I think it would go a very long way towards making the semantic model easier to set up and maintain.
Not quite right. Again, if you open another thread I can go into details and answer other questions.
The Overview page inclues three tabs: Locations, Equipment, and Properties. These are three different views of your semantic model.
On each of these tabs the main widget are List Item Cards. On the locations page the the card inclues badges so you can tell at a glance stuff like number of lights on, open doors, temp, humidity, etc.
Clicking on any individual card brings that card up where you get an ordered list of widgets showing and allowing you to control all the Points in that card. What Points are shown depends on that Point’s position in the semantic model. The widget that is shown by default for the Item is chosen based on the type and semantic model tag (e.g. a Number with the Setpoint tag will be rendered with a slider widget) and other info that’s part of the Item (e.g. State Description). The icon comes from the Category of the Item.
Clicking on the > for those widgets that have it (e.g. sensors) brings up an interaactive chart of that Item. On the Properties tab there is an “Analyze All” button which will show a chart of all the Items in that card.
All this comes for free. No configuration required beyond creating your Items and building the semantic model.
However, if the wrong widget is selected as the default, then you get into the place where you need to start customizing stuff. Here you need to add a “Default list item widget” to the Item that’s wrong. You can select from among the list of your custom widgets or you can configure a widget right there.
If the order isn’t what you want you have options to control that too. And you have some customization options for the cards (e.g. I have some custom background images on my cards.
Note, if you use a custom widget, you can then edit that widget under Developer Tools → Widgets and those edits will apply everywhere it’s used in a “custom X widget” Item metadata.
Anyway, yes, it does add finished widgets based on the properties of the Item.
Whether you like the look or not 
I spent about 20 minutes customizing mine to the point where I’m very happy with it.
The column percents are ranges. If the screen is between 0-1024 use 100%, 1025-2048 use 33% and so on. So yes, if your screen is 1023 pixels it’s going to fall back to that. If it’s 1026 pixels it’s going to jump up.
If it’s not doing this an issue needs to be filed.
I don’t understand how you get the page to fit the display part. Are you certain you are not talking about HABPanel? That’s a completely different UI and indeed, that one is not responsive. You have to build separate panels for each screen size.
Are you talking about adding the “masonary” on the page (rows and columns)? Indeed that needs to be done but it’s independent of the size of the screen.
And if you choose to use Card widgets, you can kip the masonary entirely.
My growing frustration though is that the semantic model and overview tabes are the built in way to get a usable dashboard with little to no configuration. You admit to not understand how it works but insist it doesn’t address exactly this problem. If you don’t like what’s produced that’s something that can be addressed but it strongly feels like “I don’t understand it so we must need something else”.
We already have that. That’s my point. A lot can be done to make the default look better without throwing the whole thing out and building something new.
It’s not a wizard but this is exactly what you get when:
- From the Thing’s Channels page clicking on “Add equipment to model” or “Add points to model”
Which brings up:
You can select the Location, create the Equipment Item (if you choose “Add points to model” you don’t get the option to create the Equipment and instead can choose it) and then you can select which Channels you want to include in the Equipment. Once selected you get a pre-filled out form with most of the properties of an Item that you can override as needed.
When you hit save new Items will be created for each Channel you selected already situated in the semantic model.
As discussed above, you already get a UI for the newly created Items since they are semantically tagged and this UI is the default dashboard.