the usage of ai in daily work

Stub i haven't finished writing this yet. i publish drafts early as part of WFD 17.

I use AI when writing code. Objectively it's better at writing code than doing it by hand, provided you prompt it right. Garbage in, garbage out. The quality of the output is a direct function of how well you frame the problem, the constraints, and the context. If you feed it vague instructions you get vague code.

Oxide's RFD 576 frames it well. The responsibility lies with the person in the chair. The AI is a tool. The output is yours. If the code is wrong, that's on you, because you accepted it. If the architecture is bad, that's on you, because you approved it. The moment you stop understanding what's being generated is the moment the tool becomes a liability.

the problem

Ever since WFD 15 I've been starting to wean off heavy AI usage. The reason is simple: I feel dumber.

I had a moment where I kicked off an AI task to solve something and it sat there thinking for thirty minutes. While it was thinking I already finished the task myself. I cancelled the run. Wasted tokens, wasted time, and I could've just done it from the start.

The same thing happens with explaining code I wrote. I know the codebase. I built it. But I hit mental blocks trying to articulate how something works because I've been letting AI do the tracing for me. The muscle for retracing logic and holding context in my head is atrophying.

AI is good. AI is a 10x multiplier. I've heard it all. But it's making me worse at the things I used to be good at because it's doing them for me.

the real issue

I enjoy writing code. I enjoy writing this document. Delegating everything to AI takes the fun out of the thing I actually love doing. Programming got me started because I liked solving problems. If the AI solves every problem, what's left?

arguments against reducing AI usage

• It's genuinely faster for boilerplate, tests, and repetitive work
• It catches patterns and bugs I'd miss on the first pass
• The time savings on large refactors are real
• Fighting against better tooling is a losing position long-term

arguments for reducing AI usage

• I'm losing the ability to hold complex systems in my head
• Muscle memory for debugging and tracing is degrading
• I enjoy the craft and outsourcing it defeats the point
• Over-reliance creates a dependency I don't want
• The thirty-minute-thinking incident was a wake-up call

where i currently stand

Starting to wean it off. I'm not going cold turkey, but I'm being more intentional about what I delegate and what I do myself. The boring stuff (boilerplate, repetitive refactors, test scaffolding) is fine. The interesting stuff (architecture, debugging, problem-solving) I want back.