Thank you Rob. The mistake made by those who relegate coding (and design) to grubby detail is failing to recognise that the ideas behind the design of software are more fundamental than its features. The features of software are a snapshot of the capabilities of a robust design, and in the case of a good design can change with little effort. Coding directly to features may appear to cost less in the short term but will cost more in many ways in the long term.
I read the arguments for vibe coding as arguments for that you can "produce and create without the need for caring about what you do". As though the act of caring itself would not be a central part of the process or experience, but that all that matters is frictionless flow and the volume of results. It is like an attempt to escape the fact that not only is meaningful creation difficult, but also the fact that is kinda have to be difficult, to some extend, for it to matter to us. The process of expending your energy and pushing through the friction into something that you care about is perhaps part of what makes it a worthwhile. Otherwise we might as well just be playing candy crush.
Way before the current AI coding hype he was saying how we're already relying a lot on systems we don't invest enough in understanding (or which are genuinely too complex to comprehend). When you add "vibe coding" into the mix, it does seem like a straight path to ruin, asking for more quicker without thinking about maintaining the increasingly unknowable scaffolding.
That said, we know Roman technology "collapsed" in Europe more because of the collapse of slavery system that propped it up, rather than because of the pure loss of knowledge from Romans to "barbarians" (which is also kind of dogwhistly), so it's just a directionally interesting thought, not a parallel.
“When you add ‘vibe coding’ into the mix, it does seem like a straight path to ruin…”
When I see corporations like Boeing hiring CEOs who know nothing about the company’s field and whose one area of expertise is cutting costs, it makes me think it’s inevitable that the real software engineers at those organizations will start being seen as expenses to be cut and replaced with $30K / year new college graduates with degrees in non-technical subjects who prompt ChatGPT with commands like “Write a program that flies a plane / designs a structurally sound bridge / autonomously drives a car.”
As we continue spiraling downward into a low-trust society mired in a generalized competence crisis, it will become more and more difficult to have any faith that the software underlying critically important systems was written and tested by people who know what they’re doing. All the more so when catastrophic accidents start happening regularly due to buggy, inadequately-tested auto-generated code.
Indeed the "AI" culture is finding its lows here... Contrast what we are living through with the previous artistic imaginations of "inteligent machines"; Such as the HAL computer of 2001 Space Odyssey, algorithmicly programmed with cold rational logic, would find himself embarrased by the statistical mindlessnes of the LLM.
Excellent- a really thoughtful piece that tears through the nonsense. Thank you for writing it.
Thank you Rob. The mistake made by those who relegate coding (and design) to grubby detail is failing to recognise that the ideas behind the design of software are more fundamental than its features. The features of software are a snapshot of the capabilities of a robust design, and in the case of a good design can change with little effort. Coding directly to features may appear to cost less in the short term but will cost more in many ways in the long term.
Nice writeup.
In other words, vibe coding just means acceding to mediocrity.
I read the arguments for vibe coding as arguments for that you can "produce and create without the need for caring about what you do". As though the act of caring itself would not be a central part of the process or experience, but that all that matters is frictionless flow and the volume of results. It is like an attempt to escape the fact that not only is meaningful creation difficult, but also the fact that is kinda have to be difficult, to some extend, for it to matter to us. The process of expending your energy and pushing through the friction into something that you care about is perhaps part of what makes it a worthwhile. Otherwise we might as well just be playing candy crush.
This made me think of a talk by Jonathan Blow (he's not a nice person, but has certain expertise): https://www.youtube.com/watch?v=ZSRHeXYDLko
Way before the current AI coding hype he was saying how we're already relying a lot on systems we don't invest enough in understanding (or which are genuinely too complex to comprehend). When you add "vibe coding" into the mix, it does seem like a straight path to ruin, asking for more quicker without thinking about maintaining the increasingly unknowable scaffolding.
That said, we know Roman technology "collapsed" in Europe more because of the collapse of slavery system that propped it up, rather than because of the pure loss of knowledge from Romans to "barbarians" (which is also kind of dogwhistly), so it's just a directionally interesting thought, not a parallel.
“When you add ‘vibe coding’ into the mix, it does seem like a straight path to ruin…”
When I see corporations like Boeing hiring CEOs who know nothing about the company’s field and whose one area of expertise is cutting costs, it makes me think it’s inevitable that the real software engineers at those organizations will start being seen as expenses to be cut and replaced with $30K / year new college graduates with degrees in non-technical subjects who prompt ChatGPT with commands like “Write a program that flies a plane / designs a structurally sound bridge / autonomously drives a car.”
As we continue spiraling downward into a low-trust society mired in a generalized competence crisis, it will become more and more difficult to have any faith that the software underlying critically important systems was written and tested by people who know what they’re doing. All the more so when catastrophic accidents start happening regularly due to buggy, inadequately-tested auto-generated code.
Indeed the "AI" culture is finding its lows here... Contrast what we are living through with the previous artistic imaginations of "inteligent machines"; Such as the HAL computer of 2001 Space Odyssey, algorithmicly programmed with cold rational logic, would find himself embarrased by the statistical mindlessnes of the LLM.