Consider ways to make it easier for potential users to quick start running models

[AoifeHughes]

I've been thinking about barriers to entry for new users. Here's what I'm mulling over:

When someone says "oh this sounds interesting" about TuringJl, there's quite a gap before they can actually try it out. This friction matters, especially for grabbing new users into the ecosystem.

Things maybe worth considering?

• Online environments: Are there any Google Colab-like systems for Julia? (Yes, Colab does support Julia). Ideally something ethical/free with caching support. This removes the whole "what do I need to install" barrier.
• Notebooks: I don't see them used much in Julia docs. Is this a community preference thing? They're really useful for the "here, just run this" approach - like if someone asks about a model, you can just send them a link and they're up and running. Great for prototyping before diving into a new ecosystem.
• Current tutorials: Would it be useful to have downloadable notebooks/.jl files for tutorials? Right now the most obvious path is copy-pasting into terminal/IDE which adds friction. PyMC has a nice setup where you can grab notebooks and just hit go.

Have floated this by @penelopeysm and @mhauru but thoughts / concrete imperative directions for me to implement / explore would be useful

[mhauru]

Mostly the same comments I made on Slack, with some additions:

1. I don't know of any, but see below about Pluto's instance of Binder.

2. Don't know about the community preference, but a few points that might be relevant:

• I think running .jl files in VSCode can have quite a notebook like experience. I think there’s a way to split things into “cells” in some sense and execute them individually. I haven’t tried it, but I wonder if that means that those who are familiar with the ecosystem see less of a need for notebook files.
• The Ju in Jupyter is for Julia, so if we wanted to use Jupyter that should be possible.
• There’s also Pluto.jl, which tries to be an improvement over Jupyter. I haven’t used it, but Guillaume Dalle gave a talk about it at the conference I was at just now, and I liked many of the changes in it compared to Jupyter. That could be a nice option.

I just noticed that Pluto has a Binder instance running: https://plutojl.org/#try I wonder if we could use that, or host a version of that ourselves, to allow people to run notebooks in their browser with zero installation needs? This might be an answer to question 1. too.

3. Given it’s all Quarto, I would hope there would be some way to convert all our tutorials to a more interactive notebook format. I think that could be really cool. You could read it, and just click a button and get the same thing as an editable, runnable notebook running somewhere on the cloud and accessible through your browser. Obviously also downloadable if you want to run Jupyter/Pluto locally.

In general, I agree, something like this to reduce the barrier for entry would be good. Could you:

1. See if Pluto or something else has an instance of Binder or something else Collab-like running that we could use.
2. Check if there's a known way to convert Quarto pages to notebooks of some kind, like Jupyter or Pluto. Optimally something that would integrate nicely and just give us a button like "open as notebook" on all our tutorial pages, and wouldn't require any manual effort.

[AoifeHughes]

I guess the better way to ask the question is: "What does the typical user want / use?". I'm not opposed to looking into Pluto, but I do have a strong hate for things which keep splintering off and fracturing potential user bases. Even if Jupyter has some faults if theres 10x people who use it then we should work to make that the outputs

[shravanngoswamii]

Quarto does support Jupyter notebooks https://quarto.org/docs/tools/jupyter-lab.html

[AoifeHughes]

Does it go from ipynb -> qmd or can it go in reverse too? If so, is there a nice elegant way to implement without too much change to the current setup?

[AoifeHughes]

See: #642 when it's rendered, it may partially address the above discussion.

[yebai]

Google Colab has native support for Julia now.

[penelopeysm]

Closed by #652