I think this could have been about 1/4 the length. Fairly basic stuff by modern standards and you don't need so many words. It is all correct at least, as far as I could see!

Yeah... I mean they should have just copied whatever video conferencing platforms do because they all work fine behind corporate proxies and they also don't suffer from this "increasing delay" problem.

I haven't actually looked into what they do but presumably it's something like webrtc with a fallback to HLS with closed loop feedback about the delay.

Though in fairness it doesn't sound like "watching an AI agent" is the most critical thing and mjpeg is surprisingly decent.

They use QAM and similar because it's the best way to transmit data over a small number of long wires. Exactly the opposite of wires inside a CPU.

I was looking into this recently and I didn't know this but NTFS is actually designed by competent people and is fully case sensitive.

For backwards of course Microsoft had to make the file APIs case insensitive, but the actual filesystem is case sensitive.

Also, presumably because this is a real turn-off for developers there is actually an option in Windows to sort of make specific directories case sensitive. Wild right?

https://learn.microsoft.com/en-us/windows/wsl/case-sensitivity

do people really use Lisp family for AI dev tho?

Not since the 70s. They almost all use Python.

FP has absolutely nothing to do with modern AI.

Well COBOL hasn't completely gone away... I don't think anyone expects C to become completely extinct; just very legacy.

D missed its chance. Zig is clearly going to be the successor to C, for people who don't want to use Rust.

I take time to explain why they aren’t duplicates of similar questions.

Ha yes I've found that if you explain why your question isn't a duplicate of another question and link to it, people are more likely to report it as a duplicate of that question. So stupid.

You only need to check half of the string, so I think a technically optimal solution would have some take_while and maybe char_indices in there...

LLMs don't usually make trivial language mistakes like supposively.

I think this is just a really inexperienced or unskilled developer. Or possibly a very dedicated troll.

The biggest problems with gRPC are:

1. Very complicated. Way more complexity than you want in most cases.
2. Depends on HTTP 2. I've seen people who weren't even doing web stuff reach for gRPC, and now boom you have a web server in your stack for now reason. Compare to Thrift which properly separates out encodings, transports, etc.
3. Doesn't work from the web. There are actually two modifications to gRPC to make it work on the web which means you have three different incompatible versions of gRPC with different feature sets. IIRC some of them require setting up complex proxies, some don't support streaming calls, ugh. Total mess.

Plain HTTP can be type safe. Just publish JSON schema or Typespec files or even use Protobuf.

Yeah depends on what you're using it for. CSV is terrible in many many ways but it is widely supported and much less complex.

I would guess if you're considering Parquet then your use case is probably one where you should use it.

JSON is another option, but I would only use it if you can guarantee that you'll never have more than like 100MB of data. Large JSON files are extremely painful.

I don't think I would call this functional. Python is decidedly not at all functional - there's no way to declare arbitrary functions inline, no chaining of map/filter etc.

But the static types are definitely welcome. I didn't know about the type keyword. Apparently it makes it support forward references.