AI Is Not the Solution. It's the Tool.

I keep hearing the same argument about AI that I heard about Google in 1998, and it's coming from the same kind of people.

Back then, the complaint wasn't really about search. It was about what search implied. If a machine could answer a question in half a second, what was the point of the reference librarian who'd spent a career learning where things lived? What was the point of memorizing anything, if you could just look it up? Teachers banned it from classrooms. Editors banned it from newsrooms. People wrote long, serious essays about the death of deep knowledge, about how a generation was going to grow up shallow and dependent, unable to think for themselves because a box on their desk could think for them.

They were half right and mostly wrong, and it's worth being precise about which half.

They were right that search changed what expertise meant. Knowing where information lived stopped being the valuable skill. Knowing how to ask the right question, evaluate a source, and tell a good answer from a confident-sounding bad one — that became the valuable skill, and it still is. The people who never developed it just got faster at being wrong. The people who did got enormously more capable, because the retrieval problem that used to eat a day now took thirty seconds, and that thirty seconds could go toward the part of the job that actually required a human.

They were wrong that avoiding the tool was a viable position. It wasn't a purity test. It was a bet against literacy, and it lost completely. Nobody talks about "Google-assisted research" anymore because the phrase became meaningless — it's just research now. The skill got absorbed. The people who refused to touch it on principle didn't preserve some purer form of thinking; they just did their jobs slower, with worse sourcing, for longer than they needed to.

AI is at the same inflection point, and the reaction is nearly a rerun of the same script. Substitute "hallucinate" for "unreliable results," substitute "AI-generated slop" for "linkrot and SEO spam," and you're reading the same essay with the nouns swapped.

Here's what actually matters, and it's the same thing that mattered with search: AI is not the solution to anything. It is a tool, and tools don't have opinions about what you use them for. A compiler doesn't care if you're writing something elegant or something that segfaults on line one. Search doesn't care if you're doing due diligence or falling down a conspiracy hole. AI doesn't care if you're using it to think harder or to stop thinking. The tool is neutral. What it does to your output depends entirely on what you already know, and whether you know enough to catch it when it's wrong.

That last part is the whole game, and it's where the "just prompt it and ship it" crowd and the "I refuse to touch it" crowd fail for the exact same underlying reason: neither one is doing the work of understanding what the tool actually produced. One skips the review because the output looked confident. The other skips the review because they never generated any output to review. Both end up shipping worse work than the person in the middle, who used the tool to go faster and then read every line before it went out the door.

I write infrastructure software for a living, and I use AI constantly — not to write my architecture for me, but the way I'd use a very fast, occasionally wrong junior engineer: draft this migration script, translate this error string, scaffold this boilerplate I've written forty times already and don't need to write a forty-first time. I still read all of it. I still own it. The tool didn't replace the judgment; it removed the tedious parts standing between me and the judgment. That's not a lesser way to work. It's what tools have always been for.

The people who get left behind here won't be the ones who use AI badly. Plenty of people will use it badly and still be fine, the same way plenty of people wrote bad code with a compiler and still shipped software. The people who get left behind are the ones who refuse to build any fluency with it at all, on principle, while everyone around them quietly gets faster. In five years "I don't use AI tools" will sound exactly like "I don't use search engines" sounds today: not a stance, just a gap in your skill set that somebody else doesn't have.

Learn the tool. Learn where it's strong and where it lies to you confidently. Keep the judgment yours. That's not a new lesson. We already learned it once.