I wrote and contributed to a few tech books between 2004 and 2009. One of them was obsolete before it hit the shelves, thanks to changing versions and APIs. In another case, I joined the company I'd written about and personally outmoded whole chapters. Great for my career, terrible for book sales!

This experience comes back to me as I read pieces like Dusty Phillips' My AI-Driven Identity Crisis:

So what am I good for anymore? Writing code? Vibe coding hasn’t successfully commodified software development, yet, but there is a reasonable chance it will. Writing fiction? Also not clear whether AI will take that over. Give up and focus on woodworking? Not gonna pay the bills I’ve accrued over decades of cushy software engineering salaries.

I haven't written any books since my first few, and I've often though about why.

Working on those books was a dream come true: I love writing and I love tech. Throughout my early career, I'd gotten huge value from my growing library of tech books, so contributing to that tradition felt deeply gratifying. Getting published was a solid check mark on the bucket list. The books sold okay—but publishers, unsurprisingly, are excellent at calculating advances for middling sales. I didn't expect a private island, and I didn't get one.

The real disappointment was watching my work become irrelevant so quickly. I suspect my publishing career and topic of interest hit at exactly the wrong moment: a stage where books weren't quite the right medium for fast-moving web tech.

Twenty years later, it's obvious. Back then, it felt like watching my identity shift in real-time. It's worth noting that tech books in the 2000s still came with CD-ROMs slotted into the back cover! There was even a brief discussion of whether my RSS & Atom book deserved that treatment before we settled on a ZIP file at a publisher URL.

A few years later, I worked on the Mozilla Developer Network project. I didn't write much there either, but I overhauled the content management tools for the technical writers who kept the site growing and fresh. That felt like the future: blogs, wikis, online docs, and Q&A forums had mostly eaten the space for the kind of tech books I'd expected to write.

Tech books are still written and sold, some in the same niches as mine. I don't know what the advances look like these days, but I don't buy those books anymore. I wouldn't feel the same gratification working on them.

Maybe this has all just been the tale of my own personal tech writer disillusionment and nothing bigger? But it kind of feels like I'm watching it happen again.

As Phillips notes:

But there is one successful usage of AI that is already proven: It is the go-to tool for obtaining information about technical topics. Stack Overflow usage is in absolute free-fall.

Back when I was still excited to get words on paper, Stack Overflow and the web at large were becoming the go-to tools. They ate traditional publishing's lunch. Now this new symbol manipulation and distribution technology is eating theirs.

What does this mean? I'm not sure yet. This wave of LLM tech feels more fraught than those first waves of web tech. The industry might be a bigger bubble than the dot-com bust. The models are practically built on haunted ancient burial grounds and fueled via ocean boiling.

But it's never going away entirely. Nothing is inevitable, but enough people want and are capable of progressing this technology that it will most likely progress. Kids are growing up with it now. Maybe there'll be a backlash like there was against polyester. But short of a global Butlerian Jihad, further models will be trained and this stuff will continue to work.

I've been through one cycle of watching my medium become obsolete. The question isn't whether AI will change how we create and share knowledge—it already is. The question is what we do about it.

For me, that meant pivoting from writing books to building tools that helped others publish more effectively. This time around, I'm still figuring out what the equivalent move looks like.