0x20cowboy

There is room in this world for Mcdonalds and Michelin Chefs.

I would also take your comment with a grain of salt. If the people you are talking about are javascript developers or people who do leetcode in interviews but never write that kind of code on the job, they are probably delusional. Ive seen many people call it a “craft” and what they produce is nothing of the sort.

If you really are good, you’ll still be in demand where it matters.

Those exist e.g. https://menuetos.net/ it’s just most people dont use them.

I agree with you though. If Apple open sourced os9, and the community could get it to run on modern hardware, I would run it.

I suppose it’s in how you word it. I’ve given up on trying to get a job because there is no point in trying. I can’t afford to pay my rent, but I guess you could call that early retirement.

Yeah, cool. The only way to get past resume filters now is to have AI write your CV and cover letter. TAs in school will grade you higher on your papers if you use AI so your papers sound like the median of all academic papers - you get down graded if it is "in your own voice".

So, I agree with the moral stance of the article, and I personally take all the Ls from all the CV rejection, and the lower grades because I write the paper myself. Just realize that you are choosing life on hard mode for negative societal and "dominance hierarchy" benefit.

Optimistic! There is so much to learn, invent and discover.

> There's no 'rigorous comparison' that puts CNNs over Vits

That’s not accurate. My team wrote a paper for school in which a resnet model out performed a ViT model of the same size on almost all metrics. These were smaller models, but depending on the use case that might be what you want.

I agree, but I’d extend that to any language using a package manager at this point. “A little copying is better than a little dependency” even more correct now.

All my current projects have all the code needed in the repo (unless impossible, and aside from a compiler which I guess could also be compromised)

> I’ve never understood the “if I don’t enable bad behavior, someone else will, so I might as well enable bad behavior” argument.

If someone is going to make a lot of money (fame, etc), it might as well be me.

If your codebase is pure golang it’s trivial to cross compile for different OSs (aside from binary signing on macOS). If you add some kind of C code you have to jump through a bunch of hoops to get a binary. Funny enough, I was looking for this exact project for that reason last week.

To help with any survivorship bias, 30 years in still no where close to success.

What is it?

.

Life is undefined behaviour.

I dont know what CTF stands for so I dont know if I am interested in this article or learning anything about it. Maybe I am.

Are you really arguing for not just typing out whatever 3 words this stands for once in the name of clarity?

So much this. Lisp can do things other languages have a hard time with. I think a resurgence is in order.

I wouldn’t give a junior drop access to the prod database (or anyone for that matter from a dev machine), let alone an LLM.

How do people keep doing this?

If that were true, industry interview processes would be very different.

sc has been around for quite a while: https://github.com/robrohan/sc there are several versions floating around.

“Here's the part I think a lot of people miss:”

:-/

The same argument could be made about people writing articles and influencing actions in other humans. Something, it seems, people want to use AI for. Have AI write articles for them.

Maybe have a play with them a bit more. LLMs are quite good at coding, but terrible at software engineering. You hear people talk about “guiding them” which is what I think they are getting at. You still need to know what you are doing or you’ll just drive off a cliff eventually.

At the moment I am trying to fix a vibe coded application and while each individual function is ok, the overall application is a dog’s breakfast of spaghetti which is causing many problems.

If you derive all your pleasure from actually typing the code then you’re probably toast, but if you like building whole systems (that run on production infrastructure) there is still heaps of work to do.

Not exactly what you’re saying, but a bit closer. With this library you set what css classes on the page are “hot”, it fetches the next page state and replaces that part of the page with the new state: https://github.com/robrohan/diffy

"'If I wished,' O'Brien had said, 'I could float off this floor like a soap bubble.' Winston worked it out. 'If he thinks he floats off the floor, and if I simultaneously think I see him do it, then the thing happens'".

Lol, I like that as a joke, but I wouldn’t think you are saying “a person who has no idea how something works” their opinion should be given equal weighting as someone who actually knows? Maybe you are - that seems to be how things work now.

I think you already get what I am saying, but it seems that there are maybe 3 groups. 2 who know how things work under the hood and have differing opinions and are curious to hear the other side, and one group who have no idea how things work, are very loud, have sci-fi fantasies, and spout strong opinions.

I wouldn't call that discourse i would call it ignorance.

I think that the tribalism is one sided.

On one side you have people who know how to build deep nn saying one thing, and on the other there seems to be people who don’t even know what tanh is and are very sure of their “strong” opinions.

Do you have an example of someone who actually knows how LLMs work who has a tribalistic view?

The levels would be great. I’ve tried prompting general purpose LLMs to only use words at particular levels, but they pretty quickly diverge from the prompting guardrails. One might just need to building a custom model trained only on a limited vocabulary.

I think this article misses the mark from the people in my circle.

We don’t care about “clean code” (thats mostly just juniors yak shaving into a slow system anyway). We care about correct code -code that is solving the actual problem (as close to correct as possible anyway).

Using an LLM and expecting it to be more than 60-70% is a bad idea. And using it “as a tool” - we have a hard time believing a dev reviewed, understood, and verified 130,000 line PRs every day.

Additionally, it is still unclear if the generated code violates licenses / actually “becomes” your IP.

Neat - here is another one you might find helpful (chrome only) https://cdn.robrohan.com/seq/index.html

And the source: https://github.com/robrohan/r2_seq

See https://arcprize.org/

Or visual source safe

I agree with the spirit of this, but I’d go one more step with the example:

export function clamp(value: number, min: number, max: number): number { return }

That is just adding an extra jump and entry on the callstack where you could just have done:

Math.min(Math.max(value, min), max);

Where you need it.

(The example used was probably just for illustration though)