學海無涯: learning beside a machine that learns

I learned to code with an AI beside me. Not in the marketing sense — in the literal one: most evenings, it's me, an editor, and a model that has read more code than every engineer I've ever met combined. NXTLVL wouldn't exist without that. A year ago I could not have shipped an app alone; now the distance between understanding a thing and building the thing has collapsed to almost nothing.

That collapse is genuinely wonderful. It's also worth examining, because a few of my old beliefs about learning haven't survived contact with it.

What changed

Being stuck is now optional — which is a problem. Struggle used to be the tax you paid for understanding. Now the answer is always one question away, and I've caught myself "learning" a topic in the way you can walk through a museum without seeing anything. The knowledge passes through me. I've started deliberately reintroducing friction: predicting the answer before I ask, writing the first draft of any tricky function myself, closing the assistant for one working block a day.

Taste became the bottleneck. When production is nearly free, the scarce skill is knowing what's worth producing and recognising when the output is subtly wrong. Every hour I spend with these tools convinces me more that judgment — about design, about code, about words — is the thing to invest in. The machine raises the floor; taste is still the ceiling.

Explaining is still where understanding lives. The old truth survived: I don't really know something until I can write it plainly. This logbook is partly a study technique. If an entry is muddled, my thinking was too, and no model can fix that for me.

What didn't change

The sea is still boundless. AI didn't shrink it — it gave me a faster boat, and from a faster boat you mostly discover how much more water there is. Every question I resolve while building Sparkline opens three better ones: about attention, about motivation, about what a machine should and shouldn't say to a person who is struggling to start their work.

That last question is really the heart of the studio. Tools that talk to us are becoming ordinary at astonishing speed. Whether they make us more capable and more human — or just busier and more dependent — is being decided now, in thousands of small product choices made by people like me and you.

I want to make those choices carefully, and out loud. That's the expedition.