Yes, offload your thinking to AI

I offload an enormous amount of thinking to AI. First drafts of emails I don’t want to write from scratch. Research sweeps across a topic I’m unfamiliar with. Summaries of documents I don’t have time to read fully. Rough analysis I refine rather than originate. I do this constantly, across nearly everything I work on, and I think you should too.

That sentence still makes some people uncomfortable. It sounds like an admission. Like I’m confessing to skipping the hard part and taking credit for the output. I don’t think that’s what’s happening, and I want to make the case for why, because the discomfort is based on a fear that’s older than AI and mostly wrong.

Offloading is how humans scale

We’ve been offloading thinking for as long as we’ve had tools to do it with. Writing offloaded memory. The abacus and then the calculator offloaded arithmetic. GPS offloaded spatial reasoning and route planning. Every time you delegate a task to a colleague, you’re offloading the thinking required to do it yourself. None of this is new, and none of it is inherently corrosive.

Every one of these tools got the same panic when it arrived. Calculators would rot our ability to do math. GPS would rot our sense of direction. Search engines would rot our memory because why remember anything when you can look it up. The panic shows up every time, and every time we’ve mostly absorbed the tool and moved on. The one place it landed is instructive, though: there’s decent evidence that people who lean on GPS without ever building a mental map of their city get worse at navigating without it. But the lesson from that isn’t “don’t use GPS.” It’s “know when you’re offloading in a way that atrophies a skill you still need, versus offloading in a way that frees you up for something more valuable.”

Offloading badly makes you dumb. Offloading well makes you capable of more. The difference isn’t the tool. It’s what you do with the capacity it hands back to you.

The turn

Here’s the distinction that actually matters, and it’s the one I think gets skipped in most of the “is AI making us stupid” conversations. There are two different kinds of thinking, and they’ve been affected very differently.

The first is generative thinking. Drafting. Exploring the space of possible approaches. Summarizing a pile of information into something digestible. Doing a first sweep of research to figure out what’s out there. This kind of thinking just got radically cheap. What used to take an afternoon now takes minutes, and the quality of that first pass is often genuinely good.

The second is critical thinking. Deciding whether the draft is actually good, whether the summary missed something important, whether the research sweep found the real answer or just the most confidently stated one. This kind of thinking did not get cheap. It didn’t get automated. If anything, it got more valuable, because it’s now the thing standing between you and a mountain of plausible-sounding output that may or may not be right.

That’s the trade worth naming clearly. You can offload thinking. You can’t offload critical thinking. Everything below is what that actually means in practice.

Critical thinking, unpacked

Critical thinking in the AI era isn’t one skill, it’s a chain of four, and each one depends on the one before it.

Doubt comes first, and it has to be a reflex. AI output arrives confident by default: it states a wrong number or a fabricated citation with exactly the same tone as a correct one. Your skepticism is the error correction the system doesn’t provide. Read AI output the way you’d read an unverified first draft, not a trusted expert’s finished work.

Verification is doubt made concrete. Does the number match the source? Does the summary say what the document actually said? Checking is almost always cheaper than generating, but cheaper isn’t free: you still need to know what a right answer would look like, and to actually go check. The people who get burned by AI aren’t the ones who used it. They’re the ones who skipped this step because the output looked finished.

Decision-making is where verification turns into action. AI is very good at laying out an option space, but what it hands you is a menu, not a decision. Somebody still has to choose, weighing things the model can’t see: your constraints, your risk tolerance, the thing your gut hasn’t fully articulated yet. Mistaking a well-organized menu for a decision already made is how a lot of AI-assisted work quietly goes wrong.

Accountability is why none of this is optional. Your name is on the analysis, your signature on the recommendation. “The AI said so” has never worked as a defense and never will, because nobody delegated the answerability to the AI. You can delegate labor. You can’t delegate being responsible. That’s exactly why the first three steps aren’t optional either.

Offload the labor, keep the ownership

None of this is an argument against using AI heavily. I use it heavily, in nearly every part of my work, and I’m faster and I think better for it because the generative grunt work isn’t eating my day anymore. That’s a real gain and I’m not interested in pretending otherwise out of some misplaced nostalgia for typing everything from a blank page.

But the gain only holds if you’re honest about which part you handed over. You can offload the labor of thinking. Drafting, summarizing, exploring, researching, all of it. What you can’t offload is the ownership of it: the doubt that catches what’s wrong, the verification that confirms what’s right, the decision that only you can make with context the model doesn’t have, and the accountability that was always going to land on you regardless of who did the typing.

Offload the thinking. Keep the ownership. That’s not a compromise position. That’s the whole job now.