The Problem
Alrighty, so this is one that’s been sitting in my back pocket for the past year and a half. There’s a certain “aroma” that I get when I go to vibe coded websites, or hell even use vibe coded software. And it’s not just me feeling it too, there’s a reason [AI;DR is an actual saying now](AI;DR is an actual saying nowhttps://futurism.com/artificial-intelligence/aidr-meaning). It’s because we’re all bullshitters, and you can’t bullshit a bullshitter.
I’ll follow up with a story: If I think back to the introduction of OpenClaw(then called Clawdbot when I started using it). I remember seeing the API surfaces and configurations change on an almost daily basis. So fast inf act that I just genuinely could not keep up, or care to keep up. Ultimately I just scrapped using it because it just wasn’t polished enough.
This is the type of thing that a person who has coded for 26+ years, and worked as a programmer for 15 years professionally notice as they use it. I appreciate open source software, but I don’t think I ever fully knew the reason why. I think the vibe coded generation has really taught me though.
There is a pretty stark difference between a creation, and an engineered creation. Let’s start by looking at Engineering by definition:
Engineering is the creative application of science, math, and practical experience to design, build, and improve things like machines, structures, systems, and processes.
Which is what really inspired me to write this post. I think we’re move valuable now as an engineering collective than we were three years ago. People know when they’re being served
Handling the Effects
So I think that we need to break down the effects of Vibe Coding vs. Engineering by their categories:
Continuity and Industry Standards Focus More Towards Profitability
What’s the point in writing another framework when React is so heavily ingrained into the training of models, there’s probably more ingested into an LLM about Dan Abramov & Sophie Alpert than there is about Martin Luther King Jr. at this point. This isn’t a great thing to be honest. We can already see the drift towards commercialization reaking havoc on industries.
One year ago for every major openAI or Meta release of model there was atleast 6-7 Ollama models of less than 20b, which would fit on a standard consumer graphics card. Today, you’d be hard pressed since Gemma4 got released to see any that fit on your 16GB card…and this is sad, because it means innovation has to happen in the workspace, or with those with 5k laying around. This is something we need to constantly be echoing. This keeps the power of LLMs, ML, and NLMs in the hands of billionaires where it will become more closed off from the public.
Stagnation of Free Software
I don’t mean free in the Richard Stallman sense. I mean free as in our open source licenses. We’re going to probably see a lot less programmatic growth, but using old tooling for more purposes. This is widely being accepted seeing that somehow React has become the Official Forever Framework. A lot of what we’re seeing industry wide is stagnation at the level of new technology, or the technologies that were prevalent at the time of AI adoption are becoming the primitive of their concept.
The cost of convenience(AI), is definitely going to result in a loss of innovation…so if you hated React before, you’re going to hate it for the next however long. Which leads me to my next point:
Security
It used to just be the old smelly guy at your meetup bitching about security…but now it’s real. Malicious actors ARE creative, so this new world of uncreative generation and “creation” is going to play right into their hands. We more or less will live in their playing field for a considerable amount of time. Now it’s a real threat, it can be from people somehow embedding hidden instruction in text or images via metadata, or more than likely just unicode characters. The threat vectors really aren’t changing, it’s just that the direction of the threat is moving away from humans to the agents.
Because of a lack of innovation in prevention, we’ll probably see that we are going to have problems with:
Privacy
Less people are going to engage in the public hive mind that was Steering Committees, RFCs, and other public standard bodies. No one’s got time for that when they can just ask Claude to take a rip at recreating their payroll system from the ground up.
Since we’re just really at the beginning of this phase, which is about 1.5 years since common adoption, we’re really just now starting to see the greater impact of a lot of decisions about what we’re making in the industry.
So, let’s steer this whole piece into conclusion by stating there’s a lot of optimism and growth around this industry, but there isn’t a lot of actual proof of improvement. Instead of streamlining processes we are at a point where we’re just replacing them with automation, while not replacing the space left by automation with something productive.
Frankly, to me there’s no real point if you can exploit the time savings, and just convert them into money savings in an industry like tech which relies on innovation.