The Terminal Star - tyleo.com<br>Home Blog About Me Contact Work Together Support My Work

The Terminal Star<br>2026-05-31<br>3 minute read

The dopamine machine that writes your code.

Vibe Coding without AI

When "vibe coding" entered our vocabulary it didn't strike me as something new. There have always been engineers who paper over downstream defects rather than fixing them at source. "This null check prevents the bug!" But why was the object null? Some are more interested in making pixels dance rather than understanding why they dance in the first place. "Since it fixes the bug, can't we just commit it?"

This work looks like productivity. There's evidence of 10x variance in programmer output by lines of code. But more code does not have inherent benefit. Hence the horror stories of "10x engineers" who end up creating messes instead. Speed without judgment compounds badly. "Vibe coding" just puts a name to it.

With the latest AI tools letting anyone slap 20,000-line PRs into GitHub, we can all be 10x engineers. We stare at terminals for hours on end. 3m 41s, 10.4k tokens, an animated star tickling part of our brain. We might be more productive than ever staring into that dark screen, but what is it doing to us? Are we forever more capable? Or does each prompt leave us slightly diminished?

Commodified Intelligence

As I use AI tools, I feel a vortex pulling my thinking into a simplified form. When AI works it feels like magic. When it doesn't my instinct is to slam more tokens into the machine until I get what I want. Source code flutters past and my appetite to review it shrinks. "I don't understand this plan but it seems good enough for the AI to execute." I become a vibe coder myself. I've been writing code for decades, why does it suddenly feel like a burden?

The biggest change when working with AI is a growing annoyance when we have to think for ourselves. Good engineers are replacing rigor with vibes. I don't think this is by design, it seems more like a glitch in our human programming. We pull the lever on the AI slot machine and feel a hit of dopamine. Transfixed by the terminal star, a waterfall of text drizzling down. We're elated when the output matches our intent a little sad when it does not.

✻ Welcome to Claude Code !<br>/help for help, /status for your current setup<br>✻Thinking...(0s esc to interrupt)

Does this feel good to watch?<br>When thinking is interchangeable, intelligence is a commodity. The sort of thinking we've always seen from vibe coders--pasting null checks to hide errors--has always been a commodity. The difference today is that we can generate it at scale from an addictive loop. As we push tokens into the AI dopamine machine, a faucet of cheap thinking pours out.

Don't Lose Yourself

When AI doesn't give me what I want, my instinct is to try again: more context, a better prompt, more agents, better tools. I just need to find the right words. I have 10 more minutes to spare and the machine could do the right thing this time. The desire to write the code myself feels further than ever but it may be the best move. I hone my focus, open my IDE, white-knuckle it, and get to work understanding.

To be proud of building without knowing is to be proud of being expendable. Despite proclamations that all thought will soon be mechanized, it doesn't make sense that those most capable in the future will be those most able to not think. If there is any value in original thought, the obstacle we face is to pursue it as most other thinking becomes cheap.

We now have access to commodified intelligence and we should use it, but it can't solve everything. AI taps more knowledge than ever but the dopamine-fueled mechanism biases us towards offloading thought. Our job is to understand and chart new paths forwards that can't be ordered up with commodity thinking. If we can't recognize ourselves in our work, we really are replaceable.

Back to: Blog Home

Support My Work