Faulty Towers, vibe sickness, and the vibe bobsled -- Dustycloud Brainstorms

Faulty Towers, vibe sickness, and the vibe bobsled<br>By Christine Lemmer-Webber on Fri 17 July 2026<br>I know. As if what the world needed was yet another blogpost about<br>LLMs and AI tech. Yet there is a pile of things which have been on my<br>mind, and I haven't seen them laid out elsewhere in the way I'm going<br>to write them, and so here we go.<br>I still don't use genAI to write my articles, fwiw. Here or anywhere<br>else. These rambly words are my own.<br>The tower tilts<br>I read<br>The Tower Keeps Rising<br>recently, and it has stuck in my mind.<br>The piece is an observation, and<br>according to Armin on lobste.rs,<br>it is not an advocacy for the state of affairs (though by running a<br>vibecoding company, Armin is part of advancing this direction):<br>For context: I'm the author. I intentionally did not make a<br>judgement if this is a good or bad thing, or if this is going to<br>continue working. It's primarily an observation that with agents you<br>can continue to make progress even when people on the team<br>maneuvered themselves into situations where previously they would<br>have needed to talk to each other.

The summary of Armin's post is effectively that vibecoded systems keep<br>piling code on top of code, but in many systems things seem to keep<br>building, but the abstractions keep piling on, but eventually no human<br>can understand the codebase. But this is a new way of operating,<br>because LLMs can "explain" a part of the codebase that no human can<br>make sense of, and so continue building.<br>Even if such systems continue to work, I find two things: 1) that now<br>advocates for this state of affairs have pivoted into acknowledging<br>that this is the end state of their systems and 2) they seem to be<br>accepting it as the way forward.<br>Regarding the first, I think it's very important to note that this is<br>a shift. Simon Willison, probably the<br>best pro-genAI writer on the internet (sometimes, I think, giving<br>cover for a lot of weaker writers, but is that Simon's fault?), at one<br>point coined the term<br>"agentic engineering"<br>and was very clear to<br>draw a line in the sand between agentic engineering and vibecoding:<br>We also need to read the code. My golden rule for production-quality<br>AI-assisted programming is that I won’t commit any code to my<br>repository if I couldn’t explain exactly what it does to somebody<br>else.<br>If an LLM wrote the code for you, and you then reviewed it, tested<br>it thoroughly and made sure you could explain how it works to<br>someone else that’s not vibe coding, it’s software development. The<br>usage of an LLM to support that activity is immaterial.

In an incredibly short period of time, basically a year, Simon<br>changed published a fairly honest article titled<br>Vibe coding and agentic engineering are getting closer than I’d like:<br>The problem is that as the coding agents get more reliable, I’m not<br>reviewing every line of code that they write anymore, even for my<br>production level stuff.<br>I know full well that if you ask Claude Code to build a JSON API<br>endpoint that runs a SQL query and outputs the results as JSON, it’s<br>just going to do it right. It’s not going to mess that up. You have<br>it add automated tests, you have it add documentation, you know it’s<br>going to be good.<br>But I’m not reviewing that code. And now I’ve got that feeling of<br>guilt: if I haven’t reviewed the code, is it really responsible for<br>me to use this in production?

It's a great read, and what I will say is that I applaud Simon's<br>honesty and willingness to self-reflect and challenge prior<br>statements.<br>But the gap of time between the<br>former and<br>latter<br>articles are stunningly short, just slightly over a year.<br>And Simon isn't alone. Just a year ago, I think the memetic shape<br>was by and large that something along the lines of "agentic<br>engineering" is what people could or should do, and, though I think<br>many people are hesitant to admit it, I think most people using these<br>tools are tending towards vibecoding and not agentic engineering, just<br>as Simon himself found himself pulled.<br>Before we look at the consequences to this, I think we should look at<br>why it's happening.<br>The vibe bobsled<br>As far as I know I'm the only person who uses the term "vibe bobsled"<br>and, well, I doubt it's a term that's particularly likely to catch on,<br>but I find it personally useful.<br>Bobsledding, if you are<br>unaware, is a particularly strange and interesting sport. It's a lot<br>of fun, but you don't have a lot of agency in it. You sit in a<br>bobsled, you go down an icy track, and really, there is only one way<br>to go. But people can become experts in it, and can indeed measure<br>themselves against each others skills; it's an olympic sport, and I<br>remember my own first encounters with bobsledding as a child, when my<br>father and uncle and aunts took me, and it was thrilling like a roller<br>coaster and intoxicating upon my first encounter.<br>But again, ultimately, there's only one place to go.<br>The vehicle is the LLM, you are the passenger. And I think the amount<br>of agency...