[tda]

Ome advantage of python is that it is so slow that if you choose the wrong algorithm or data structure that soon gets obvious. And for complicated stuff this is exactly where I find the LLMs struggle. So I make a first version in Python, and only when I am happy with the results and the speed feels reasonable compared to the problem complexity, I ask Claude Code to port the critical parts to Rust.

[rabisg]

The last part is really interesting. It feels like the whole world will soon become Python/JS because thats what LLMs are good at. Very few people will then take the pain of optimizing it

[eru]

The LLMs are pretty good at optimising.

Not because they are brilliant, but because they are pretty good at throwing pretty much all known techniques at a problem. And they also don't tire of profiling and running experiments.

[miki123211]

If there's one thing LLMs are really, really good at, it's having a target and then hitting / improving upon that target.

If you have a comprehensive test suite or a realistic benchmark, saying "make tests pass" or "make benchmark go up" works wonders.

LLMs are really good at knowing patterns, we still need programmers to know which pattern to apply when. We'll soon reach a point where you'll be able to say "X is slow, do autoresearch on X" and X will just magically get faster.

The reason we can't yet isn't because LLMs are stupid, it's because autoresearch is a relatively new (last month or so) concept and hasn't yet entered into LLM pretraining corpora. LLMs can already do this, you just need to be a little bit more explicit in explaining exactly what you need them to do.

[eru]

> The reason we can't yet isn't because LLMs are stupid, it's because autoresearch is a relatively new (last month or so) concept [...]

I'm not so sure. People have been doing stuff like (hyper) parameter search for ages. And profiling and trying out lots of things systematically has been the go-to approach for performance optimisation since forever; making an LLM instead of a human do that is the obvious thing to try?

The concept of 'autoresearch' might bring with it some interesting and useful new wrinkles, but on a fundamental level it's not rocket science.

[philipallstar]

I've not tried this yet, but doesn't it use up loads of tokens? How do you do it efficiently?

[eru]

It uses a lot of minutes on your computer(s), since you need to run lots and lots of experiments.

I'm not sure if it's particularly token hungry.

[elcritch]

Not just profiling, but decoding protocols too.

Recently I tried Codex/GPT5 with updating a bluetooth library for batteries and it was able to start capturing bluetooth packets and comparing them with the libraries other models. It was indefatigable. I didn't even know if was so easy to capture BLE packets.

[eru]

Could you ask the LLM to do a write-up on the process and post it? (Or you can write a blog post by hand. Like a caveman. ;)

[elcritch]

I find writing by hand is the best. LLMs spit out such linked-in writing that I don’t even want to read it. ;)

But that would be a good blog post and I got some travel coming up. But honestly it was just “oh here’s a BLE python library, see if we can get it running”. I prefer Codex because it seems to do well for guiding the LLMs for complete engineering changes.

[anthk]

Wireshark would do that. But you need to understand low level tools because in case on some BGP attack you all LLM developers will be fired in the spot.

Flakey internet connection: most of current 'soy devs' would be useless. Even more with boosted up chatbots.

[eru]

> Flakey internet connection: most of current 'soy devs' would be useless.

We used to make the same jokes about Googling Stackoverflow since before many users on this site were born.

[anthk]

And it's partially true. Offline documentation should be mandatory everywhere. Networks can be degraded tomorrow in the current 2nd Cold War we are living. And, yes, the states and goverments have private backbones for the military/academia/healthcare and so on, but the rest it's screwed.

When the blackout the only protocols which worked fine where IRC, Gopher and Gemini. I could resort to using IRC->Bitlbee to chat against different people of the world, read news and proxy web sites over Gemini (the proto, not the shitty AI). But, for the rest, the average folk? half an our to fetch a non-working page.

That with a newspaper, go figure with the rest. And today tons of projects use sites with tons of JS and unnecesary trackers and data. In case of a small BGP attack, most projects done with LLM's will be damned. Because they won't even have experience on coding without LLM's. Without docs it's game over.

Also tons of languages pull dependencies. Linux distros with tons of DVD's can survive offline with Python, but good luck deploying NPM, Python and the rest projects to different OSes. If you are lucky you can resort to the bundled Go dependencies in Debian and cross compile, and the same with MinGW cross compiling against Windows with some Win32, SDL, DX support but that's it.

With QT Creator and MinGW, well, yes, you could build something reliable enough -being cross platform- and with Lazarus/Free Pascal, but forget about current projects downloading 20000 dependencies.

[mirsadm]

Not in my experience. They're pretty good at getting average performance which is often better than most programmers seem to be willing to aim for.

[eru]

What kind of 'average' is this, if it's better than what seems to be typical?

[9rx]

> JS because thats what LLMs are good at.

That has not been my experience. JS/TS requires the most hand-holding, by far. LLMs are no doubt assumed to be good at JS due to the sheer amount of training data, but a lot of those inputs are of really poor quality, and even among the high quality inputs there isn't a whole lot of consistency in how they are written. That seems to trip up the LLMs. If anything, LLMs might finally be what breaks the JS camel's back. Although browser dominance still makes that unlikely.

> Very few people will then take the pain of optimizing it

Today's LLMs rarely take the initiative to write benchmarks, but if you ask it will and then will iterate on optimizing using the benchmark results as feedback. It works fairly well. There is a conceivable near future where LLMs or LLM tools will start doing this automatically.

[rabisg]

My experience is from trying to get the React Native example to work with OpenUI. Felt Sonnet/Opus was much better at figuring out whats wrong with the current React implementation and fixing it than it was with React Native

But yes I see what you mean and I think people are trying to solve it with skills and harnesses at the application layer but its not there yet

[deterministic]

Nope. The world runs on code written in C and C++. Including Python itself. There is a reason why there are literally millions of C/C++ programmers out there working on C/C++ code every day.