I've actually outsourced the first iteration of that to AI. :-)

At the end of coding sessions, I generall ask it to spin up an adversial agent whose task it is to find flaws, inconsistencies, mistakes, bugs, etc.

I'd still strongly recommend to anyone shipping AI-generated code to have a thorough human code review, but for a first pass at surface level, the AI can check itself if it has some kind of separation between the generator and the reviewer.

Personally, after some experimenting, I've concluded that:

• AI is good at implementing known algorithms.
• AI is good in creating code that flows straightforward from a design document.
• AI is good at code reviews and finding bugs.
• AI (unlike me) doesn't mind writing a few hundred tests.
• But, it tends to over-engineering simple tasks to the point of "why does Hello World need enterprise-software levels of abstraction?"
• It struggles with code that doesn't have a known solution.
• It's difficult to get it into a different lane once it started going down the wrong way.

What I use it for these days is rapid prototyping. Using AI I can make a functioning prototype of an idea I have in a few days instead of a few weeks. I can quickly test out an idea for a feature or change in my existing codebase.

What I also have been using it for a while is as a code-reviewer and for someone like me who is mostly working solo, that's an insanely valuable thing. Not just the bugs themselves, but also the ability to discuss code with someone. Basically, AI is a rubber ducky that replies.

Mickeysoft lost the plot again.

The game-changing element of LLMs and agents is that they have a non-computer-language interface - you can talk to them in plain English (or Chinese or whatever). This makes complex computer interactions available to a much wider audience than any previous computing invention. Adapting humans to the AI is getting that backwards. It's a sign you don't understand the first thing about why AI is so widely successful.

I predict this will be an abject failure and a year from now will quietly be pushed to the very back of the bottom-most drawer, never to be spoken off again (though it is getting crowded in there for M$).

Meanwhile, in the real world, everyone I've talked to about these things already has a skills file to teach the AI about their personal coding style. This is what actual humans came up with, independently several times, and it works: Adapt the AI to you. Plus we've been writing coding style guides for humans for a long time now, we've got plenty of experience doing that at least somewhat right.

No, that's when you want a RAID system with multiple disks.

This is only for small corruption, like a failing drive might cause, or due to cosmic rays/bitflips.

This is just EC for a non-redundant drive.

That's such a dumb, smoothbrained description I doubted the capabilities of the person who wrote it - at first.

Then I realized he was right. There isn't a filesystem that does RS or, really, any sort of single-device EC. This is a genuinely good idea, at least for spinning disks, where sector corruption is more likely.

This may be personal anecdote only, but I've only seen 1 SSD get any sort of data corruption as a failure mode, and it was a very early iteration - maybe a 8 or 16GB one. SSDs seem to simply fail outright - the FRU or something else failing for one reason or another. The storage is fine.

Who's going to be able to scrounge up an extra 12KB of RAM in today's market?

The size in metric would be about 2.5 kilometres.

That would have been really useful for some mag tape systems I used to use. They had too many errors for their error correction system.

OTOH, this kind of thing is very likely to have bugs. Test it THOROUGLY before folks start depending on it. Perhaps, after it passes initial tests, make it an optional choice in the installer, and require extra steps to do the install (with warnings: CAUTION: BETA FILE SYSTEM HAS BEEN CHOSEN).

File systems need more thorough testing that most software choices.

Currently a good rule of thumb is "If it's easy to check the answer relative to creating a good answer, AI is a good choice.". That's just a rule of thumb (i.e. it's not always true and use sense in applying it), and that's just "this was true up until this month from at least the start of the year". There's no guarantee that it will be true next month, but that's the way to bet. Next year gets more problematic.

OTOH, robots will introduce an entirely new set of problems. But this rule of thumb is pretty likely to be good for them when they start being significant.

I don't.

The creator of the pi harness did say that the claude system prompt changes often enough that he had to build his own tooling to try and improve reliability and repeatability.

I'm not convinced the differences in roles I've observed are claude personality differences as much as humans giving different prompts.

Yeah, it *is* better, if you can afford it. Just don't forget that it's dangerous. Care is needed. Possibly even extreme care.

A point, but that's a place the "selection" part of evolution comes in.

I said "good method". There are lots of ways to estimate things. And using the best methods for cells, some cells will be estimated older than other cells in the same body. Many estimates rely on epigenetic markers, but those can be removed either artificially or by normal processes...but that doesn't make the cell actually younger, and here by "younger" I mean "more error free" (i.e. more similar to the initial cell from which the body was generated [and "similar" is tricky to define in this context]), what it does is allow cells that have been marked "dubious" for one reason or another to begin reproducing again. DNA copying is NOT perfect. Each cell in your body is most likely mutated from the original cell. Most (surviving) mutants are harmless, but some aren't. So cells that have reproduced the "estimated maximum safe number of times" stop dividing. (There are exceptions, like the cells linking the intestines, but those are kept to as minimal a number as is feasible.)

I had just mentioned the reason:
Cells tend to accumulate mutations. This is especially true each time they divide. And this is well known. If your evolutionary line didn't include a such a limit, you would likely not be around due to any of a number of problems, cancer among them.

There is no good marker for the "age" of a cell. Mutations tend to accumulate over life, and there's a reason for the Hayflick limit (even if it isn't universal).

Manipulating cells so that they will continue dividing is dangerous to whoever gets them. It may be necessary, but don't fool yourself about its danger. This sounds like a recipe for brain cancer.