No chance. Who knows what sort of other impact it has on the nervous system. Also, imagine the "enshittified" version of this product.

"most compromised packages are taken down quickly" -> because people notice it right away. Set a minimum age and people won't notice it early.

I have explored in #dotnet Using ClrMD for string analysis

Unless you want to get your hands into memory dumps.

I have explored Object Stack Allocation in .NET 10

I agree

So this is interesting. I high stake organizations the code is reviewed by multiple eyes principal before being merged. The reviewers are equally responsible. But key difference is negligence and bad intentions. The question then: in what use-cases using unsupervised AI means negligence?

I have explored SIMD Sum 2

Laszlo - .NET Developer, Personal Blog In this post, I explore a couple of ways to create *lazy properties* in C# and .NET 10. What does a *lazy property* mean in the context of this post? It is an object instance property that gets initia...

Creating lazy properties with .NET10

#dotnet #csharp

blog.ladeak.net/posts/lazy-p...

Things must improve.

Which of these languages are best for AI? A rhetorical poll.

New article: Which of these languages are best for AI? blog.ploeh.dk/2026/03/09/w...

I had a long chat about this with friends. If nobody reads the code anyway, or not needed to touch it, I suppose it could be assembly (or a platform independent IL). If *I* am expected to optimize or fix parts of AI generated code (now or later) then the language *I* am the most fluent with.

I know these tools, and I am thinking buying a hw for it, but not sure what could be the best, given there is everything from a few hundred euros to thousands. Models are getting smaller and cheaper to run.

Do you have sources in this space to follow? I am really keen on having something locally.

So that to have AI to review AI written code, or simply for yourself?

But I don't get how law works. Some things are considered the same but some not. ie. the Uber vs Taxi.

These are typically not regulated markets (incl. crypto). They are prone the all the market manipulations that we have seen in the past. That is why they are popular: people try to get rich by using old tricks on unregulated markets and naive users.

Because there is a package for everything, and most of the work is boring integration and glue of packages.

I also notice that reviewing code is hard. In teams I worked only 1 out of 10 devs can do a decent code review. I think this is the main problem for long term maintainabilty, and circles back the question. Does the language matter, why not just assembly code then?

My experience with AI is that it generates OK code. On my primary language I can still write significantly better performing code. I also often find that it generates a solution, but as soon as I see it, I realize how to improve it. Something, that I would notice otherwise much sooner if I write it

Why does AI still write code on a certain language? Does the programming language matter? If it generates overblown solutions over time that nobody reads anyway, why not generate assembly or some sort of a cross platform IL and simplify the compilation step?

got it, thanks

Why not back up to a blob in azure storage?

I think re-usable building blocks are readily available more than ever before. Most enterprise "projects" are just integrating services in the simplest possible way without any real added value. 0 engineering, 0 thinking. This does not require talent.

Is thinking about it as a blog tailored CMS a good analogy? CMS like features make it valuable compared to say static blog site generation?

Yes, two tasks that are processed one after the other on single CPU, or concurrently (competing) on 2 CPUs.

Ambient contexts are usually culture, timezones, etc., but can be also CPU count, internet speed.

Markdig Markdig A fast, powerful, CommonMark compliant, extensible Markdown processor for .NET. Get started Documentation GitHub dotnet add package Markdig Available on NuGet — .NET Standard 2.0+ Playground Try Markdig live with the public API. Edit Markdown, pick extensions, and click Run (or press Ctrl+Enter). Markdown inputThe public API truncates input to the first 1000 characters. Extensionsadvanced common common+pipetables+tasklists+footnotes advanced+nohtml common+autoidentifiers+mathematics This uses...

And last but not least, I have promoted Markdig to 1.0 🎉

This project has been around for 10 years and to celebrate this, the project has now a proper website with its documentation https://xoofx.github.io/markdig/ 🤩

AI inverse:
The difference between a GUI and a TUI is the difference between arranging a room by simply moving the furniture where you want it, versus trying to decorate by typing out strict X and Y coordinates for every chair.

This is happening to me too. It is harder to reason about methods because they get (or not) inlined, and suddenly that completely changes the perf characteristics. Add something, IL gets bigger and not inlined anymore. It seems to have become more sensitive on things we cannot control easily.