I saw today that my CoCalc (https://cocalc.com/) license was about to renew. I've been meaning to move to a more #SelfHosting or #P2P paradigm, so this was good timing for me to cancel. I left this message when doing so:

"I want to support Sage, and I like using CoCalc, but I don't want to put money towards a service that supports Microsoft, Amazon, or OpenAI in any way. I didn't care so much about this before, and I even log in to CoCalc with a GitHub account, but I'm tired of having my work scraped for profit by people who donate to fascists who are destroying my nation. I would consider returning if all ties to these corporations were cut and cloud computing services came from responsible companies, perhaps in the EU."

I'm pretty sure I already have backups of everything I want from GitHub, so I can be done with them too. I should have quit when Microsoft first arrived. My plan is to switch to Radicle (https://radicle.xyz/) rather than another centralized service. I'll be sure to post about how that goes.

I've been a fan of Sage (https://www.sagemath.org/) and CoCalc (https://cocalc.com/) for some time, and I would now like to complain.

Sage is an open source computer algebra system which is written in Python, but for technical reasons comes with its own version of the Python interpreter. The first IDE I used regularly was Eclipse, and I used to know the arcane steps needed to make it use the Sage Python binary on various systems.

I switched to PyCharm some years ago, but I have only been using it to write pure Python. Now I want to use Sage, so I tried doing the same gymnastics I used to do with Eclipse and was annoyed.

I decided that being forced to work online wasn't a big deal, so I'd use CoCalc instead. Even though CoCalc was made with Sage in mind, there does not seem to be direct support for running Python modules with Sage, only the notebook style is promoted. It seems absurd to me that after all these years it is still so much work to simply use Sage in a Python project with an IDE.

In addition to this, Sage's support for the kinds of calculations I want to do at the moment is quite immature, with TODO in many, many places.

I want to use Sage since it combines many useful (and fast) libraries that I need, but I think I need to just accept that I should start building my own solution that works for my purposes.

I know that I could contribute to Sage, but I feel like the weight of changing what I need to change is so much, and that I would get my calculation done faster by just doing it myself.

I discovered today that #ChatGPT and #CoCalc (which already has AI integration) make it extremely easy to perform any #sagemath computation that can be described to GPT; I used this to perform enough numerics to arrive at the solution to a problem in https://mathoverflow.net/a/454051/766 . (The GPT provided code did contain some minor syntax errors, but CoCalc’s native AI could easily fix them.) I did not feel proficient enough in the past to use Sage on a regular basis, but now I think I will.

So cocalc.com now has ChatGPT built in. You tell it in english what you want it to do, and it will give you the code. It makes errors, and what I just tried "Make a graph of sin(x) and several taylor polynomials" that I needed for a class demonstration. It made a pretty good go at it, though it wasn't completely legal code, but easy to fix. This is genuinely useful, a great starting point to whatever you are trying to do.
#math #sagemath #cocalc #chatgpt

Here I'm trying to write notes in #cocalc / #sage, in this case about @johncarlosbaez and James Dolan chatting about cyclotomic fields:

https://math.robert-figura.de/share/public_paths/43337889eb6ae74925f438e586f0b9025a2d6079

My notes aren't complete, and it turns out that I'm very much not a fan of MathJax notation.

For my taste I'm fighting too much with the user interface, It's much better than cocalc from a few months ago, and I suppose it just takes some more practice to make it bearable.

Anyways, it is nice to have symbolic calculation available.

Does anyone know if #CoCalc can have a Jupyter notebook kernel running in the background?

Experimenting with NNs and training is taking time that the connection is broken before completing.

I just tried #CoCalc yesterday for the first time, and I really like it. I should have tried it sooner.