Goodbye 2024, Welcome 2025

With only a few hours remaining of 2024, it is time to look back and take a tiny outlook on 2025.

Migrating to .Net 8

A lot was going on at work. The new management brought a lot of changes, and many more are yet to come. It definitely does not get boring for the foreseeable future.

While the move from .Net 4.8 (Framework) to .Net 6 (Core) was a lot of work, the migration from .Net 6 to .Net 8 was much simpler. It still offered a few quirks, but nothing compared to the move to .Net Core.

With our code generators and helpers, we could move our last large application from .Net 4.8 directly to .Net 8. While a few tests are still outstanding, we eagerly await the push to production.

A separate blog for Python Friday

With the upcoming 250th post of Python Friday, I thought it would be a good opportunity to move the series on its own blog. I wanted something that lets me write the posts in Markdown and publish it with ease. I explored a few options and ended up with Material for MkDocs as the platform. After the technical part was done, the real work started, and I had to move 250 posts to Markdown. While there are helpful tools to do the migration, checking all posts and fixing formatting errors took its time. But now everything runs smoothly on PythonFriday.dev, and I am happy with the result.

Moving away from WordPress

After the successful migration of Python Friday, I noticed more and more things that I dislike on WordPress. At the same time, the mud-slinging between WordPress.com and WP Engine got into the spotlight. The arguments made there by WordPress.com pushed away the last doubts and made me plan the migration of my WordPress blogs.

In December I moved the 460 posts for Improve & Repeat to MkDocs. I could apply all my knowledge from the first migration and got it done in about the same time as it took to move Python Friday.

My old blog GraberJ with 200 posts was hosted on WordPress.com. The migration there was a pain, then I could not use any of the batch tools to fix the tags and categories. Even worse was the plug-in for source code, which was such a pain that every post with code needed a manual fix. In the end, this too could be solved, and the blog archive is now accessible on jgraber.ch/blog_archiv/.

The great northeastern adventure

My holiday this year brought me a nice mix of old and new places. I started in Oslo after the NDC and drove all the way up to the Nordkap. It was impressive to see how much changed in the 10 years since I was the last time up that far nord. From there, the journey continued through Finland, the Baltic countries and Poland before I went back to Switzerland.

You can find the pictures I took on Flickr, where you also can find my smaller journeys in 2024:

• Copenhagen
• Barcelona
• The Netherlands & Belgium

Advent of Code

I looked forward to participating again in Advent of Code. Unfortunately, I had a lot going on and had to follow a strict timebox to solve the problems. While I did not enjoy the story as much as the one from last year, many of the puzzles where a lot more to my liking. I noticed the gain in confidence to solve the puzzles in Python and when I was stuck, I found help to not only solve the specific problem, but also to improve my knowledge of Python.

With the introduction of the free tier of GitHub Copilot on 18. December, I eagerly tried it on the remaining puzzles. While the first attempt worked like a charm, the other puzzles quickly showed the limitations of this AI tool. A few puzzles took many iterations to get the correct result, while other puzzles could not be solved at all.

Long story short: Even with extensively described problems, GitHub Copilot may miss the important parts of the task. While it creates the glue code fast and in a usable form, it made a mess with the algorithmic part. There is a lot of room for improvement, and I am not convinced that the current approach will be able to fix it. We will see.

2025?

As far as I can tell, there will be no shortage of work for 2025. Lots of ideas fly around and if we want to realise them, it will fill the next year without any problems.

I expect a massive push to AI in 2025. Not necessarily because the tools are ready, but because big companies put so much money into it that it now must show some results it their revenue. It is therefore time to gain some hands-on experience building LLMs and machine learning projects and not just using a finished product. That way I can better understand what is possible, what of that makes sense for a given problem and where the marketing crap starts.

I wish you a happy 2025!