We’ve opened signups on Nextjournal today and would love to get your feedback: https://nextjournal.com/mk/public-beta.
Nextjournal is a multi-language notebook platform with first class support for Julia (as well as Python, R, Clojure).
The hosted notebook space is getting a bit crowded, but I believe we offer some unique features that are missing in alternatives:
- any software and data is backed into a sharable docker container, which greatly improves reproducibility and makes it easy to create base images for other users
- you can remix any article, which automatically makes all used data & software available in your new article
- publishing a notebook as a runnable and polished article is one click away
- you can share notebooks with peers and invite them to collaborate on the same notebook in real-time
- it’s not based on Jupyter, but one can import Jupyter notebooks and should have all the same features available
- One can import & export articles as Markdown and all docker images are downloadable, so nobody needs to be afraid to get locked in
- it’s free for open science (and will stay free) and costs money for private articles and research groups. We’re still figuring out the exact terms, but I’m sure we’ll find a good model for everyone.
For the lazy, a quick walk-through:
Some example articles:
https://nextjournal.com/sdanisch/julia-gpu-programming
https://nextjournal.com/r3tex/loss-landscape
https://nextjournal.com/sdanisch/makie-show
https://nextjournal.com/sdanisch/webio-jscall
https://nextjournal.com/julia/billiard
Now, to play around with e.g. Flux and the FashionMNIST data-set, you should be able to just remix the loss-landscape article, which already has these packages and datasets downloaded, installed and precompiled!
Let me know if you have any questions!
Best,
Simon