Does code readability still matter?

I work as a software developer on scientific applications. These days, a lot of the code I write is generated by Claude. When I first started using AI, I would read and understand pretty much every line before committing. Nowadays, I mostly look at the code, check that the overall logic makes sense, and then move on. However, I often get comments from colleagues on the PRs along the lines of: “Is this code readable?” or “Would someone picking this up in a year be able to understand it?” Honestly, the answer is mostly likely no. This makes me wonder how much we should still care about code being readable to humans. If an LLM is writing and maintaining most of the code, perhaps it eventually makes more sense to optimize for code that is easy for an LLM to understand rather than a human. I also wonder whether we will still be reading code a year from now. What is your take on this?