AI is the hottest topic right now. Everyone has opinions on it. Some call it the future, others decry it. I fall closer, but not fully, into the latter camp.

In the following I will try to formulate my thoughts on the current state of AI.

The tech is amazing. Just considering AI understanding the input. Consider how primitive the old text-based adventure games had to parse user input. That alone is an impressive feat. And that’s not what everyone is excited about.

Furthermore, it is a novel way to interact with a big chunk of knowledge on a surface level.

But I have limited my AI usage.

AI for Coding #

For coding, I use Supermaven in Neovim as a fancy autocomplete – never generating more than one line at a time. It is a very convenient way to write Go error messages.

AI seems to be “okay” for web development tasks, particularly frontend, from what I can see online. In other fields the penetration seems lower.

I’ve tried to utilize AI with Ansible. For simple tasks or roles it works nicely. But I had to intervene manually for more complex tasks to fix things it messed up. That costed me time.

I can see the potential to replace the boring refactoring tasks (that can’t be done with even a complex search-replace) or writing boilerplate code.

AI for the Creative Arts #

Generating funny images to share with friends seems to be the largest use case so far. But the idea of commercializing that feels off and given OpenAI’s recent shutdown of Sora, the industry might see it the same?

I generally don’t understand why it was pushed so hard in this direction.

I’ve used image generation occasionally to create inside jokes for my friends. My friends and I did use some AI music generation tool to make some stupid songs as well. But both are just for fun in our friend group, and we would not be sad if we could not have done that.

AI for (Re)search #

One of the original big goals of AI from (I think) OpenAI was to cure cancer. I don’t know how that is going, but it seems not a goal any more. At least no one is talking about it.

I do use AI as a replacement for many web searches. This might be a larger indicator for the quality decline of modern web search engines and the messed up world of SEO – in large part fuelled by AI slop websites.

AI in other Fields #

AI is used in many more fields. But I do not have any experience in those. Making it hard to judge.

Early experiments to replace human customer service people with AI have been unsuccessful from what I’ve read.

So it appears to me that outsourcing jobs to AI is not yet feasible.

Financial Situation #

The finances around AI are crazy. So much money. How are they ever going to make profits to pay that back? I understand that the goal is to just build AGI and use it to generate all the profits. But the behaviour of these companies does not fill me with confidence that they will achieve AGI anytime soon. And in that case, Nvidia is going to end up the only lasting winner. As they were after the crypto hype.

I will not buy an Nvidia GPU as long as Jensen stays CEO. He has shown his colours.

Fear of Replacement #

Currently, I’m not afraid of being replaced by AI. It is true that there have been layoffs in software development that have been attributed to AI. This sucks, of course. But I’m unsure if there were no other reasons for layoffs. I can see a regularity in the layoffs of large tech companies since the covid pandemic.

My Ideal Future #

I have recently started to explore AI locally on my machine with Qwen 3.5. Of course, I’m still figuring things out, but so far I’m impressed how well this works on my consumer hardware (AMD Radeon RX 9070 XT).

I like to believe that local AI could be the future. Companies should push to reduce the hardware requirements of their models and release them into the public. I’m not asking them to do that for free. They could release a smaller model for free personal use and the larger ones under a commercial licence.

From what I’ve seen in the tech world, innovative ideas are often coming from individuals trying things out.