While the world is busy debating who really solved Navier-Stokes, or discussing how many years are left for humanity, I have a much bigger problem: I’m bored.

I’m (as of now, but how long still?) a software engineer and my wife is one too. We often discuss how our job has completely changed and how it keeps evolving. I was pretty skeptic on LLMs up to last year, but the progress since then is pretty evident. Our typical day has gone from writing/reviewing code, finding bugs, discussing software architecture etc. to instead constantly prompting LLMs and waiting for their answers. At first I didn’t give up, trying to still do everything myself and only resorting to query the robots when I felt I needed a second opinion or help in exploring APIs/libraries which I wasn’t familiar with. Now that has changed as well, models are constantly improving and I recon a good one can do most probably a better job than me in a much smaller amount of time. What I’m left with is a little bit of excitement whenever I found the model didn’t catch an edge case (probably because my prompt wasn’t good enough) or wrote something overly complicated for a straightforward task, but other than that I feel pretty much demotivated.

I have always been curious and eager to look at bugs in a codebase, and I truly enjoyed the process: looking into logs, reconstructing the sequence that led a particular error, trying to reproduce it, having that “aha” moment when finally the cause is clear and the fix too. The more something would be difficult to uncover/fix the better would be the satisfaction at the end. Now all of that is gone: a bug pops up, and: “ok dear robot can you please look into it?” I’ll be waiting for the next 5 minutes while I get bored watching videos or reading articles on the web. There’s no incentive at all for me in spending time on working on the bug myself, it would probably take longer anyway. At the same time I still have to keep monitoring the robots somehow, so that I’m ready to feed the next prompt to the machine. So I’m stuck in this limbo where my attention is half focused on the next prompt and half on whatever meaningless thing I’m doing on the side. It’s like being a passenger in a car that you can’t drive yourself, but you still have to give directions to the driver and make sure they don’t crash. While being bored is certainly bad, I think the fact that we’re not spending time deeply thinking about problems is even worse. What can we do to fix this?

I don’t know the solution, but I gather this is a common problem now (first software engineers, more recently mathematicians. Who is next?). A few days ago I stumbled upon this blog post and it resonated a lot with what I’ve been thinking about. By handing over everything to LLMs, we will eventually never “develop an intuition for how systems behave and fail”. The post draws a comparison to air traffic incidents: they’re incredibly rare, but when they happen the pilots need to be ready on how to face the situation. Currently, in a software company where everyone is using LLMs to code, it would be incredibly counterproductive to not do it. However, one still needs to not only develop (I hope university is still useful on that side) but also maintain a certain level of critical thinking/craftmanship. “You might pick up a few things from watching Serena Williams play, but you only learn tennis by getting on the court”. The post is about (software) incidents and how to deal with them, but it can be applied to anything. We have to come to terms with the fact that most problems will be solved by AI soon. That doesn’t mean we should stop trying ourselves. Also, solving problems is fun. It’s not just about the final solution, but more about the process to get there. Similar concepts were expressed by Douglas Hofstadter in an article I recently read. This one is about the process of learning languages and enjoying it. But again, it can be applied to anything.

Having conversations in fluent French, acquired by dint of immense hard work over many years, was an unmatchably rewarding experience.

…

“I have a serum I can give you via injection, and once it’s in your bloodstream, your Chinese will be as good as a native speaker’s. Want it?” I would answer, “No, thanks.” I want to try to conquer the Everest of the Chinese language on my own. I don’t want a helicopter ride to the summit.

So what can we do to fight the boredom and cerebral atrophy? I recently rediscovered that algorithms are fun! I’ve been very excited recently finding out that a problem at work (thanks Claude!) could be solved by a variant of the line-breaking algorithm from Knuth-Plass (very popular on hackernews as well - see https://news.ycombinator.com/item?id=28537923 and https://news.ycombinator.com/item?id=34689213)! Now I have another problem that could be modeled as an uncapacitated facility location problem (on non-metric spaces) and I’m enjoying exploring all the different ways it could be solved, again with Claude as a great partner. Having to think about algorithms is a nice distraction (and I’m not the only one who thinks that), but probably not the ultimate solution. For now though I don’t have better ones, so I will probably try to go back to writing algorithms to solve small defined problems. I see many people share the same feelings (see https://news.ycombinator.com/item?id=49854875 or https://blog.ploeh.dk/2026/09/16/on-learning-programming-in-an-age-of-llms/) and as we say in Italian “mal comune, mezzo gaudio” (that could be roughly translated as “Misery loves company” according to ChatGPT), at least we’re not alone in this journey!