Danke für den Tipp 😊

A tragedy in 3 acts.

2026-03-05 Why I Don’t Vibe Code

Like many developers, I have a whole folder of draft hobby projects that have never been finished. […] From the LLM perspective, these are folders of failures and I could indeed use LLM to make an app a day or whatever challenge I want. However, the process was far more important than the product (again!). Not every whimsy needs to become a reality. Often, I get more from the fun of brainstorming and the process of learning enough to know that I don’t need to continue and finish the job.

2026-04-29 The LLM Is Not a Junior Engineer

Contrary to how I might seem at times, I do think there could be a place for LLM models in modern software development. I do understand why some developers are entranced with the ability to create applications in hours and turn prototypes into products. I see how it could be highly useful for home hobbyists, back-office bureaucrats and expert engineers to build the low-stakes software they’ve never had the time or skill for.

2026-10-25 My Local LLM Stack

And yet, as a developer, I do feel like I need to work with these tools to understand their potential for augmenting my own software development work. There are some cases (basically as a limited coding assistant) where a LLM can be useful. Some recent examples I have used across several projects:

• Look through this codebase and find all the areas where I have # FIXME comments. Collect them in a report and analyze if any no longer apply.
• Improve the test coverage for this file to be above 90%
• Update the openapi spec for this project in the repo to match what the API actually implements today […]
• Look through the data here and find all cases where there are likely errors. For instance, if a person has an end date before their start date.

This reminded me of @rhamph’s post last week. Is this an archetype of guy that is concious of the harms of LLMs, but cannot trust his own gut enough to see they are fundamentally wrong? So they try to come up with their own super-ethical spin on the beast and try to change the rotting carcass from within. But was anything gained here? I feel that an ideal of simple, elegant, and empowering code was lost.

deleted by creator

You know that people in the Sovier Union also faced consequences if they didn’t work, and did in fact get paid wages, right?

What makes people put in effort is a complicated mix of material, social and psychological incentives, and I don’t think sweeping generalizations like that are very helpful to understand anything.

Was für ein selbst-Bericht, das erinnert mich an ein echtes Nestlé Plakat das ich mal gesehen hab:

WIR WOLLEN DASS KINDER ARBEITEN

Irgend ein Stock Foto von Kindern hier

AN IHREN HAUSAUFGABEN

It’s quite something. Found deep in the discussion section:

Inserts ignored provided columns, creating rows best described as “deeply cursed” ¹²

12. Jepsen encourages the consistency model research community to come up with more official terms for this.

Thanks! Well generating the summary would have the downside (for me), that I would have to look at it.

But the semantic search bit was the one thing I felt was an actual new capability. I’m trying to imagine what if Google tried to pick this as their new moat (a better but resource-intensive way to search, scaling benefits if you establish yourself as the Internet’s main vector store, people might actually pay for queries…), but no they joined the team that wants so shove slop down everybody’s eyeballs.

Can you answer a question about RAG for me? My understanding is based entirely on this diagram from Wikipedia

If RAG takes into account vector similarity between query and reference material, do you need an LLM to regurgitate the found best vectors back to you? Could you get back a list of documents that best matched your query, or are the results in an inscrutable LLM-internal representation that need to be transformered back into readable text?

Depressingly plausible, thanks!

Looks like instead they will commit their capacity to AI In Space and a little Starlink (so probably mostly Starlink, even if Elon currently drinks his own Cool-Aid).

Really drives it home for me how institutionally cucked the US gov is. For this to be as big a deal, no-one must have the slightest concern that the US will claw back or conscript SpaceX capacity for national security. And correctly so.

I didn’t expect a Mastodon web client with monospace font could go as hard as it does. It goes so well with the cyberpunk theme.

The magazine published a reader’s response to this piece, that (seems to me, a total amateur) dismantles its economic claims with basic Marx. But they don’t go into the “interminable detours” about the weird stuff.

You don’t know how ready I was to receive the most toxic reply ever conceived by man.

I declare your takes to be extremely good and correct and, frankly, you’re incredibly cool and attractive for having them.

I’ve been enjoying the writing of Zoe Baker recently. Specifically Means and Ends about the historical origins of anarchism. While looking for that, I also saw an essay on Anarchism and Democracy that references more recent thinkers. Might check that out too.

Anyway, one of the takeaways for me was this: the fundamental opposition to parlamentary democracy entered the (european) anarchist canon in a time, when firstly parliaments were typically created as concessions by monarchs to stave off revolution, and by design ineffectual. And secondly violent revolution was still very much on the table. In this context participation in state democracy was opposed categorically as a strategic error.

If you live in a country where these two facts are no longer accurate (okay #1 is debatable, but #2 generally no), the strategic conclusions should also be different. Even if you have a post-state vision of democracy, some level of involvement in state democracy (or at least not torpedoing those efforts on principle) is non-optional in my opinion. Not least because fascists are trying to capture those systems to suppress all of our movements even harder.

Banger post to start a blog.

Link seems borked. Fixed version: https://ti-kallisti.com/infosec/tales/vibes.html

Benn Jordan posted a mega-update video on his epic battles against Flock.

His incredible work doesn’t quite take the sting off my disappointment in his decision to use an AI-generated image at 18:30. (Come on, the option to badly crop your face onto the “This is fine.” dog was right there!)

Edit1: oof, it’s not a single incident. 20:30 now. I hope it’s just for this one sub-story, or I might not care to finish…

Edit2: It was kept in check afterward, though I suspect it was used for some assets.

I suppose this could make sense if you think about people worried about climate/ecosystem collapse, and the nuclear standoff before that, and maybe the rise of totalitarianism and mechanized warfare before that…

But that ignores the vast host of batshit doomsday prophets that were around long before and all throughout that part of history.

Fine sneer by data journalist and senior engineer Jacob Harris: Why I Don’t Vibe Code

(slightly dated (Mar 2026) but first time I heard of this guy)

Is anyone familiar with the website I Can Eat Glass? It was apparently created in the 90s by a Harvard student, who was concerned that people learning foreign languages only learn common phrases (e.g. “Where’s the loo?”) which makes them stand out as tourists when they speak. If one were to say a truly uncommon phrase, it stands to reason one would be instantly respected as a native. So he set up a website to collect as many translations of the phrase “I can eat glass, it does not hurt me” as he could.

Surely you are riveted to your screens right now, desperate to learn what this very special brain genius has been up to since then. Ethan Mollick is a professor at a Pennsylvania business school and published “Co-Intelligence: Living and Working with AI” in 2024. Some say he is substacking about the exponential growth of AI capabilities to this day, but critically this book was apparently read by some german teaching influencer.

He in turn published a self-help book for parents of school kids this year, that got put on display in my local library. And of course it has a chapter on AI, so I had to lend it and hate-read a little. And of course he references Mollick and his Co-Intelligence screed, which he apparently absorbed as a base expectation what the future of AI will look like, and this is how I got to learn about one very special brain genius from Harvard.