Westerhof

With 2026 on the treshold, I wanted to reflect on this year together with the 2025 look-ahead post and see what was promised, what was accomplished and what I missed the mark on.

Do more with AI

I’ve messed around with AI this year, but it was limited to small-time problem solving with ChatGPT. Think small script functions and getting solutions to compiler errors in my code. Most of my dedicated LLM usage was it’s integration into my Home Asisstance instance, helping me control parts of my home or for being a conversationalists.

I’m not bullish nor bearish on AI, but must admit a deep hope that the AI hype dies in 2026. The impact of the industry on consumer hardware intensifies my beliefs. Next to that, any AI integration I’ve had the pleasure to experience professionally was incredibly half-baked and the results on User Experience were terrible. I continue to be flabbergasted that during any Proof of Concept we do, I keep hearing project members being surprised by the non-deterministic aspect of any LLM. Hearing “It does something different even while I give it the same input” continues to make me chuckle. The layman expects that with the LLM, dumb computers are now smart.

They’re not (yet), just 80%.

Hearing those that I respect and who work (adjacent to) software explain happily how to use PLAN files to iterate on LLM output for over 10x times, doing review after review on the generated code, makes me cringe. Specially because I know they can do much better. THey are intelligent, talented and skilled people that I have seen master difficult concepts and solve difficult problems without AI. They don’t need it, and I would fear them losing this mastery because they’re outsourcing their thinking. One can absolutely argue that this is great for regular people to be able to ‘program’, to get on their level. But would you let people that don’t have field experience work at that level, with at an unprecedented layer of abstraction?

Sigh. I digress.

Much like the dot com bubble, I hope that whatever remains after the fallout offers a great experience to customers and buckets of second-hand enterprise hardware for prosumers.

Release a game

This promise I’ve kept somewhat. While I did release a post about creating a mod, I did not release a game on it’s own.

It’s not for not trying. I’ve spent a couple of months learning Unity by creating a clone of the infamous Banana clicker game from steam, which was even more barebones back then. I’ve also tried to creating an online version of Pazaak from KOTOR using Golang and Websockets.

There’s still plans for me to release both, just for the experience.

Next to that, as a side hobby, I’ve also worked multiple holidays on creating a RenPy visual novel. Here the writing came naturally as I could ferry most of the experience from mod creation and hosting Dungeons and Dragons sesisons. But creating the assets took on significatnt time and is the reason it is not published. Wanting to creating animations for certain scenes was the biggest hurdle here. Sure, the dompanine reward for seeing what you animate ‘come to life’ in front of you is great, but working in Blender remains a headscratcher.

All-in-all I’m saddened that I’ve not been able to release a product. I love playing video games and it scratches my creative side when working on one and would have loved to be able to share something more tangible. On the upside, I notice my programming skills increase a lot even while attempting to make one. My Pazaak attempt in particular: working on a Golang that’s adjecent to, but works completely different than my REST API helped me understand the language way better. The refresher on making use of websockets was good as well, something I had last tinkered with in University.

So if you’re reading this, and are technically inclined, I recommend you make a game for next year as well. And it does not matter what kind of game or if you release it.

Getting certified

In 2025, I set out to get 2 certifications: OSCP and CISSP. Instead, I’ve ended up with one: an OSCP lookalike called eCPPT.

I some regret not fully completing this goal. But the moment I had my mouse hovering over the ‘Buy now!’ button, going for OSCP felt very daunting for me. The idea of a proctured exam was not something I felt comfortable with at the time and I had in the mindset of failure at the time. I felt much more comfortable going for eCPPT. Reasons being that it was cheaper and lowered my sense of guilt, the reviews made it appear more ‘industry friendly’ and, most importantly, an experienced collegue of mine is very positive about the certification.

That pushed me to go for it starting in October. 3 long months later, I started the 24 hour exam on the 21st of December and passed with flying colours. I felt suprisingly anxious about it, fearing my Active Directroy skills were not up to par for me to absolutely p0wn the corpo network. But, thanks to my notes and some extra guides online, I was able to pass it.

Expect a post about it next year, because we will:

Blog more

Over the course of 2025, I wrote a total of 5 posts. That includes this one and the original 2025 post. My plan included to write a minimum of 12 posts, one per month. Next to that, while I do have all the subjects and drafts lined up, I simply did not finish them.

This is simply because I had no time to write anything with the depth I wanted it to be. The main reason I was even able to write this post and the 2026 look-ahead post was that an increase in me having to do multi-hour traveling with public transportation. Writing in an empty(ish) train very nice and is something I predict is going to happen much more often next year.

Because of this, I’m certain next year’s goal will be met - but that’s for the 2026 plans post!

What else happened?

Goals aside, 2025 was a good year for me. Some other reflections:

• Started contributing to Ukraine’s war against the invader both in time, money and resources. If eager and able, you can do so yourself as well here at United24, or if you have a 3D printer, consider joining DrukArmy.

• Despite sounding like corpo-speak, I’ve gotten a lot of satisfaction from multiple talks and projects at work being finished. The collaboration with my peers gave me a great sense of accomplishment. A good counter to imposter syndrome!

• My artist profile keeps slowly gaining reach. Despite only releasing 13 posts with a pattern that can be plotted on an exponential decay graph, my follower count has grown 31%. I’ve scheduled and buffered pieces for periodic release next year and am looking forward to seeing the continued growth. There’s still significant progress and quick wins to be made, but I’m very contend. Besides that it earns me no money, the community feedback has made it worthwhile.

• My friends and family are healthier, we laugh and spend more time together doing things we enjoy. My collegues are still great to work with and I value their insights, feedback and smiles. As such, a good life from dusk till dawn!

On to next year!

Changelog

• 2025/12/26 - Included eCPPT outcome + finial
• 2025/12/18 - Reviewed
• 2025/12/04 - Initial draft