A smart take that's half right
A prominent founder / investor recently wrote:
"when engineers use AI to code, you pay frontier prices to make the codebase bigger, while the engineer learns nothing."
Sharp sounding quote, but it's not correct.
It reveals not only a lack of current knowledge, but also assumes there's only one way to use the tools: the naive way.
For instance, give two people an oven: one burns toast, and the other bakes a soufflé. Same oven, but different operators.
One label applied to an entire field
We took Karpathy's "vibe coding" term and bolted it onto every use of AI in development. For anyone whose mental model froze in Q1 2025, that single label is the whole map, and many let that shortcut define the whole space.
Sure, you can flip on "dangerously skip permissions," write a clever prompt, and YOLO an app without ever looking at code. Fine for prototypes and throwaways. That's not what serious people are doing.
Agentic-assisted engineering is a completely different beast. It requires multiple planning & architecture phases, detailed specifications, acceptance criteria, context management, and dozens of small decisions per session. The result is that an engineer finishes more literate in the system than they started.
The same intellectual move is everywhere
The "vibe coding" moniker also has many cousins that share the same trait: vocabulary that froze before the field caught up to itself.
"AI hallucinates" gets repeated by people who've never heard of context engineering. "I just need better prompts" comes from the same camp. And "AI makes us dumber" is the laziest take in the catalog, as it treats every human as identically passive.
I'm sure in the 70s there was a fear that handing someone a calculator meant no one would do arithmetic anymore; however it allowed us to put rockets in space and balance portfolios in real time.
The true problem is that these folks haven't traveled along the AI Adoption Continuum. In that model, they're still on “bicycles”, while some are zipping around in “spaceships”.
The conversation can't happen across the gap
If you've never heard of a “spaceship”, it’s nearly impossible for me to describe one. You might even conclude, from your own experience, that personal transport offers modest upside. That's coherent reasoning from inside your frame.
I called this the Time Traveler's Dilemma:
You can't explain the future to someone whose entire frame of reference is the past.
Here's what I've noticed: these takes often come from people who are consuming, not building. They're reading consulting decks, listening to podcasts, and attending conferences, instead of actually putting hands on the metal. It's easier to recycle comfortable takes than to understand the nuances of how the field is moving.
In Time Traveler's Dilemma terms, they're sitting in the hot tub expecting to be carried through time by proximity. Meanwhile, those of us who've put in our Gladwellian 10,000 hours building & operating are the ones driving the proverbial DeLorean.
The insidious effect
The worst part is that bad vocabulary gives smart people permission to opt out. To the untrained ear, it sounds like craftsmanship and rigor. When it's really Dunning-Kruger dressed up as rationality.
The damage isn't individual, as the blast radius extends to teams and organizations. Every week spent arguing "vibe coding won't work for this project" is a week the real conversation never happens. And that gap only widens as everyone else’s knowledge compounds.
Labels are the rate limit on progress. Name the craft, or stay stuck arguing about the shortcut.
