As an early adopter of llms I am very happy with my job right now. I feel like my job becomes easier over time as llms can do more stuff. They take over more and more of my programming tasks. The clients and most other developers do not yet realise the extent to which it’s possible to hand over complex tasks to the llm. They typically don’t realise how to use them properly and discard them as a gimmick.

The problem is that this create a false sense of security. The reality is that my job will get harder over time as more and more developers become familiar with llms. They will become better developers and the competition will increase a lot. Clients will know what’s possible and will call bullshit if a task takes too long.

On top of that, the llms themselves will start taking over tasks as “agents” (aka for loops). They will get assigned a goal and just work towards that goal. Especially when there is a proper bridge between browser / the Internet and llms in terms of vision and actions. But this will come sooner than we think.

The comparison of “just another tool in our toolbox” or “we just move up one abstraction layer in the stack” stops being true when the tool itself can reason and can directly compete with you.

I suspect it will take longer than most people think though to replace software engineers. There are a lot of inefficiencies that need to be solved. But it will be a very messy process, with big chunks being taken over at once while other areas are lagging a lot.

If I had to give advice to younger people, I’d say go into a very “human” career path where your charisma and charm goes a long way. Or start a business. I don’t see llms automate “making money” right away. It’s messy. There’s no overarching directive for the llm to follow, at least not one that it can break down into subtasks and start chipping away at it. Finding product market fit is a very messy, real-world process with feedback loops from humans. Another option is going into the field of AI and become the engineer working on the actual llms. I’m an ML engineer but I mean go deeper than what I’m doing. Really understand the nitty gritty details and become one of the few people that can train and finetune huge models. And of course, start investing aggressively.

As for me? I’m trying to start a business by creating something new every week in 2025. Maybe something will come off it. I’m trying to change the incentive structure of my life because right now it’s not good. I’m also trying to invest the money I have. And I keep learning about ML, but I can’t seem to muster the courage to delve into the real hardcore maths / technical stuff. At least not yet. I know quite a bit about it but I’ve never worked for a business where I could actually work on really big models. Who knows one day…

—

This is my third blog post and also my third “creation” for 2025 out of the total of 52.

I could’t find the willpower to create something more complex like an actual project.

In fact, I almost gave up on the entire endeavour like I typically do with these types of things. My thinking almost immediately went to “no-one will want to read this shit, why would I write anything”.

But I’m doing this for me. Done is better than perfect. So I’m just pushing this article online anyway.