2025: The year LLMs felt real

As a normal, non-technical person, I think we are living in magical times—and I say this because of LLMs, especially the LLM coding tools. As someone who doesn’t know a lick of programming, the fact that a person with absolutely no idea what coding is can still wrangle these tools and get useful things built is the very definition of magic to me.

I also think the discourse about AI has become a little weird. These tools can be phenomenal and genuinely useful, and I feel like that basic fact needs to be revisited over and over again—because most of the debates about AI (not that they aren’t important) end up being unhelpful at an individual level.

To my mind, 2025 has been a breakout year for large language models. Progress continues apace, and these things are now at a stage where we can all have our own Jarvis in our pocket. I say this because I just finished reading Simon Willison’s fantastic roundup—2025: The year in LLMs—and it’s packed with insight.

Even though I’m non-technical, Simon Willison’s blog is one of my favorite places to keep up with what’s actually happening in AI. And a big reason it stays so good is his post on link-blogging: posts and links, consistently, with taste. He also publishes via Substack—so subscribe, add his RSS feed, and make sure you read his blog.