AI as a Work Partner

When ChatGPT first showed up, I treated it like a magic trick. I'd throw prompts at it to see what it could do, marvel for a minute, and move on. Somewhere in the year and a half since, that changed without my really noticing. On my own projects, on my own time, these tools stopped being a novelty and became part of how I actually work. Not a thing I test. A thing I reach for.

How the Work Changed

The clearest shift is the blank page. Drafting anything, a design doc for a side project, an outline, a tricky email, used to start with staring at nothing. Now I start with a rough draft the model produced in seconds and spend my time editing instead of generating. The first version is rarely good. It is almost always a useful place to start.

Coding is the other big one. For my own builds, a coding assistant handles the boilerplate and the parts where I half-remember the syntax, while I stay focused on how the thing should be structured. It is faster, but more than that, it keeps me in the flow of the problem instead of breaking off to go look things up.

Research and learning compressed too. When I want to understand something new, I can interrogate a model, ask follow-ups, get a wrong answer and push back on it, and reach a working understanding faster than I would piecing it together from scattered sources. The catch is that I have to know enough to catch it when it is confidently wrong, which it still is, often.

Copilot, Not Autopilot

The mindset that makes this work is simple. The model accelerates my work. It does not do the thinking for me. When I have tried to hand it the whole job and walk away, the result is mediocre and sometimes wrong in ways that would embarrass me if I shipped it. When I treat it as a fast, tireless collaborator that I direct and correct, it makes me genuinely faster and sometimes better.

That distinction matters more than it sounds. The judgment, the taste, the sense of whether an answer is actually right, that still has to come from me. The tool raises my ceiling. It does not hand me the judgment to use it well. If anything, it rewards knowing the domain more than ever, because the person who can tell good output from plausible-looking nonsense gets enormous leverage, and the person who cannot gets confidently misled.

Why I'm Putting in the Reps

None of this touches my actual job in any direct way, and that is deliberate. There is no version of the current moment where these tools belong anywhere near regulated work without a governance framework that does not yet exist. Building that framework is careful work, and it should come before any deployment, not after.

So the reps are personal. I am building the muscle now, on my own time, because understanding a tool from daily use is a different thing from understanding it from a demo. When the policy and the guardrails catch up, and in a regulated industry they will, on the careful timeline that work deserves, I want to already know how this technology behaves, where it helps, and where it quietly leads you wrong.

What It Comes To

AI has become a real work partner. Not the oracle the hype promised, and not something that produces anything worthwhile without a capable person steering it. It is a fast, able collaborator that makes someone who knows what they are doing meaningfully more productive. The value scales with the judgment of the person using it. These tools raise the ceiling, but somebody still has to know where the ceiling should be.