A: Need a separate topic area for talking about the Web interface to ChatGPT compared to the API and programming.
B: Canvas usability.
At present canvas is not usable. I spend an hour last night trying to delete a canvas file, jumping through hoops with my assistant. I still haven’t deleted it. I was also unsuccessful at any attempt to edit it.
I observed something called library
in my sidebar. Initially CHAT said that my canvases should be individually listed and should have 3 dot ‘more’ menu where an option was for delete. Not present.
If I click on Library, I get thumbnails of my canvases.
Clicking on any one gets me a larger view, with the option to edit the image.
Chat said there should be a 3 dot ‘more’ menu with an option to delete. Not Present.
If I click “edit image” I get a prompt to describe the changes. I don’t have an option to change anything myself. When I hit enter, I get a prompt that says, "There are a lot of images being processed, please wait. I have given this up to overnight, and it does not resolve. Reloading the page works.
I typed the request: DELETE THIS CANVAS. at the prompt. Did not work.
I think taht these are not canvas documents but they are images. I don’t want images, I want an editable document.
I do NOT want to create an image by description and multi-hour waits. I want to be able to interact directly with the picture. I don’t require that this needs to be Adobe Illustrator, but it should be, say a notch above Microsoft Paint. Something like Geogebra workbooks, combined with SVG, MD, flowchart symbols/shapes that you could assign meanings to, arrows, lines, tags.
Most of the interaction of this should be a local app running on my PC/Mac/Ipad, with bundled messages to CHAT regarding updates/changes on some kind of basis.
Interaction needs have the potential for real time. E.g: Sort highighted items by count of events. Add tag #foo to blue box. (One of the things that was great about NeXTStep is that many things could be done in multiiple ways, reducing restrictions on workflows)
Coupled with this, Canvas needs multiple ways to export:
- A well defined format for other application developers to create import filters for.
- As Encapsulated Postscript – better for placing in a document.
- As common raster files at least JPEG, and PNG.
- As PDF – better for page oriented documents.
- As HTML5 / SVG – for online use.
- As MS Word – tricky.
- As Adobe Illustrator. – this would be a subset of Illustrator’s full capability. Need to figure out how to do round trip. Off the top of my head, given Illustrator’s capability for layers, CHAT could add a layer or layer group that held the metadata that it needed after a user modified the data.