I think it's an optical illusion. It looks weird to me, but comparing it with various objects I have I do feel like they're parallel.

Specifically the roof edge vs base edge, the back of the roof vs the front of the roof.

The most amazing thing about this isn't that ownership and lines of responsibility and politics are messed up, it's that the guy holding the reins is 100% aware of this and has no power to get the horse to move in the right direction. Is there not an org chart? How is there not a person above who knows what who below them owns? If these are the execs below Bill, how does Bill not know which one owns it? If he knows and they're not taking responsibility, why doesn't he fire them? Reading this I'm honestly more surprised that Microsoft lasted past 2004.

I don't think GP was claiming they weren't.

I clicked on that expecting to see something about the limitations of retained mode, but I didn't see anything. There's a claim about long term retained mode complexity, and a discussion about a winit bug...

Doesn't the placeholder imply nothingness? E.g. 10 means 1 ten and 0 ones.

No, that implication is not automatic. The Babylonian numeral system evolved over milenaries, and the simple space noting the absence of a digit between two digits was later noted by a "0" digit. But this symbol was never used by itself, so it did not mean "nothingness". In the same way, a civilization having the concepts of addition and subtraction does not automatically induce the concept of negative numbers.

On the other side, ancient Egyptian numerals did have a zero, but it was only for nothingness. But they did not have a placeholder zero, because their systems was similar to Greek and Roman numerals.

That doesn't solve the problem of it hallucinating more design details and putting them in comments. The solution to too much authoritative documentation isn't more authoritative documentation.

Comments exist for humans to understand the why of the code, including perhaps external dependencies/assumptions/requirements.

LLMs inserting code comments makes zero sense by definition.

That’s not true because LLMs clearly benefit from having multiple statements representing aspects of the same thing.

That’s why they’re always doing the “it’s not X, it’s Y” kind of patterns. It’s reinforcement.

LLMs benefit from knowing the why, the intent, and the reason/justification for code to exist, too.

Catching code desyncing from it is a valuable place to reconsider assumptions and maybe even invariants.

> LLMs inserting code comments makes zero sense by definition.

Not completely, because anyone who reviews the code in the future does not have access to your original prompt, so theoretically leaving a comment that explains the "why" portion of the prompt would be useful. LLMs rarely do that, though.

If secrecy doesn't matter, why did they tell the agent not to divulge stuff in the first place?

I posted this a couple days ago and I guess it got re-raised. I thought it was cool -

1. I didn't know there was any static analysis systems out there with type info, it's famously missing from codegen etc.

2. Most of the discussions are about formal analysis, kani etc. But I really haven't seen much of more traditional (?) analysis in rust

3. It can identify unnecessary `clone` (with some restrictions, probably) which was always something that bothered me... whether to clone or not is almost purely mechanical but you end up having to choose by hand.

Isn't this like self-stimming or self-regulatory behavior?

URLs are URIs and URLs include "how to access this" which fundamentally makes it hard to prevent them from changing. If it were an abstract ID, like a UUID, or a hash or something I'd get it. So are all URLs bad URIs?

The "R" in URL is a misnomer. Designers of the early internet really wanted internet to be made of documents. It turned out the internet is actually made of interactive apps and services. Hence the impedance mismatch between terms and their practical meaning. URL was supposed to be used to locate resources in the resouce-oriented internet - instead, in the service-oriented internet, URL is a language of sorts for making remote procedure calls, some of which are read-only operations returning resources. So URLs are only as stable as the RPCs powering the service behind it. As services evolve, so do URLs.

That's why browsers need to change "bookmark" to also create a local snapshot of the page (on every visit).