I am so fucking glad i pivoted to teaching. Now I run a Makerspace and teach young nerds creative problem solving. The pay is less but every day is a joy.
The whoop band confirms it
don’t be the condom. be defective.
look whos the monkey on a type writer now :D
I’m unemployed, I’m incredibly snarky at the recruitment posts. (This literally hasn’t changed in decades.) Right now I’m like “oh, they want Python programmers…? Good, I’m one. For web scraping tasks…? Okay that’s bullshit. Wait, this is an AI job post? The MBAs, drunk with their Vibe Coding power, thought they could feed the Machine, but lo, they were mistaken!!! For web scraping was a Dark Art.”
Me, a programmer, without any formal education, eating popcorn and reading the comments.
Just a little bit of impostor syndrome tonight boys, feelin’ good otherwise :DNot even. Your job isn’t to protect prod, it’s to take the blame when prod goes down.
aint being the scapegoat exactly how you protect prod?
If they expect you to write or go through 1000s of lines of code in a single day. your whole job isn’t programming, is to be the scapegoat.
If I can’t understand the code, and you can’t explain, I’m not validating your pull request.
End of story.
Just ask ChatGPT to explain the code
It’s great at that, had a colleague run a diff through Claude and it explained how the new code was better and how the optimizations worked. Then my colleague realised they’d run the diff the wrong way round, fed Claude the new diff and Claude happily explained how the new code was better and how the optimizations worked.
Ah the ol please
Also called a reverse centaur.
I’m realizing how over engineered Claude’s solutions are. I have realized now just today that it’s best at just dding tiny updates to code. Not good at architecting anything new at all that has to work for longer than 5 days.
I spent like an hour yesterday bullying Claude out of using a ResizeObserver, Refs and a useEffect construct just to fix the height of a container (there was a bug where the container didn’t properly scale down in height because some weird hard-coded properties in an external component inside of it). I knew that it had to be solvable with just the right css selector, but I was lazy and didn’t want to manually go through the html and figure out which selector on which component would be the right one. And on paper “Go through this massive block of html and check which property is out of place” should be the kind of use case where LLMs actually work. The stupid thing introduced two new different catastrophic failure modes where the entire browser crashed due to excessive rerenders before it found something that even remotely worked. And then I had to spend another hour stripping out unnecessary style properties it glued on while looking for a working solution. In the end, I probably would have been much faster doing it by hand.
I threw together a slop project to test something. It put in some JSON for something that didn’t use that. I told it it was just some plain text. It took the JSON library out again. Then I added in something that does use JSON and it tried to make a basic JSON parsing library from scratch.
It’s like a very junior programmer at times, with amnesia.
It does make some menial tasks simpler, but if people think it will actually replace us then they’re in for a shock.
It does make some menial tasks simpler, but if people think it will actually replace us then they’re in for a shock.
That’s the sales pitch, we’re in a job market drought right now. Give it a year or two and all the slop will begin breaking, making a new job market boom for developers
But honestly I don’t want to spend my time fixing vibe coded stuff.
Would be kinda cool if we could get People Who Know What They’re Doing™ to just…rebuild stuff from the ground up, untainted by slop. Haha a guy can dream.
yeah imagine that, hiring experts to excise their expertise…
what silly thought pained laughter
Yea if you don’t explicitly give it direction on how to keep the code maintainable, it will gladly just spew shit into the same file until you have 10K LOC in it.
Claude make me an app that makes me $1 Million/day. Do not make any mistakes. Make the code maintain able.
Thanks for the tips /s
To be fair I used to make entire features like this and then get R# to move the classes into separate files at the end. Guess I don’t like to organise too early, before the code has found it’s true form?
Remember that you are also collecting the responsibility of 10 and the pay of 1.
🫠 too real
Shift + Tab + Fuck it.
I think the pace of development is such that it’s only a few years before humans are no longer in the chain of decision making. It was just 2017 when the infamous transformer paper came out.
A stupid person would believe this.
“AI” in its current iteration is a distillation of human knowledge. It lowers the barrier to access to long standing programming libraries that make people think it is some sort of smart. It seems intelligent because it’s imitating the things we have already created.
it’s imitating the things we have already created.
dont you think thats exactly what most people want? they dont want to create next Linux :(
Its more how unbelievably wasteful “AI” is to give dumbshits access to very basic tools. If and when it gets more efficient it will be viewed as we see search engines.
There was literally the dark version of Google to save energy on CRTs, and now it’s energy consumption is low and normalized.
Yeah, that’s what bothers me so much. It doesn’t get “smarter” or imagine novel solutions. It simply has access to everything that’s already existed.
I’m scared we’re
gonna havealready experiencing some kind of intellectual slump because handing everything to the machine built on existing knowledge…doesn’t yield any new ideas.I’m sure they’re still banking on human brains having new ideas, so they can simply steal them and feed those new ideas to the machine. Then they’ll tell the ideator their brain is irrelevant thanks to the industrialized intellect-theft device.
The intellectual divide will be deepened significantly by “AI”.
Ive never heard dumber conversations than people thinking “AI” is solving some hard problem. Here are a few examples:
- It pulls text out of a pdf
- It can read data out of a dwg
- It can overlay data into a 3d model
- It can update configuration data across lots of code
All of these could easily be done before, people were just too stupid to know how to do it without typing a stupid prompt and wasting resources to do it.
I wonder if this whole phenomenon could be explained by a Dunning-Kruger type effect. Where dumb people see “AI” as a great solution, and those on the smarter side see straight through it as a loud over juiced parrot.
No need to wonder. You are spot on.
LLMs are incapable of reasoning. They just guess the next token.
Neurons are incapable of reasoning they’re just firing based on a deterministic action potential. What are neural networks doing but creating internal models of reality and guessing what’s next based on patterns?
We would’ve bullied those nerds more for misusing the word neuron. It was cute at the time, but now they’re running around being all weird thinking that since they can pretend to read that one paper a better way to do one computer thing, it makes them experts in how brain works, just because there are similar words.
I don’t think anyone who is aware enough is under the assumption that neural nets are a one to one replication of neurons. It’s more of a special type of graph. That said one could also represent a human brain as a special type of graph.
The amount of good computer people specialised in ml, who think this knowledge gives them understanding of human mind that even neurobioligists don’t posess is staggering. And depressing.
Also not really the point I was arguing. It’s pretty clear NN are very different than a human brain. It does though give us some insight to minds more generally but I don’t think current networks are on the order of complexity necessary to really be generating abstractions similar to a human neural network. That said there’s strong arguments to be made about glimpses here and there like models spontaneously creating their own internal workspace like an inner monologue or quasi subconscious. To be clear I’m not saying it is aware or identical to a human mind but it’s interesting to see it self assemlble similar looking structures without any direct guidance.
LLMs do not fit as neurons in your comparison.
Brains can reason, LLMs cannot.
I’m not saying they are identical but they are similar in the sense that a neuron cannot reason. It’s only in a quite large and densly connected neural network we believe there is reasoning.
But the closest equivalent of a neuron in the AI world would be a node, and I haven’t been talking about that. So your comparison does not fit or make sense.
LLMs as a genre of AI research are incapable of reasoning. At best they can pretend to reason, and even then they are terrible at it.
I think both can more aptly be represented as a special type of graph like from graph theory. Once again fundamentally the human brain isn’t reasoning on the level of neurons and most definitely atoms aren’t reasoning
What is the transformer paper
Worst people you know pretend they read a math paper from 10 years ago, and it gives them deep understanding of all the complex problems you can imagine. It’s hard to overstate how much this one paper fried their brains. It’s even harder to overstate how it’s actually just one paper about computer math
The white paper on the fundamentals of the current AI push. Everything is for the most part based on that architecture from 2017 but with cleaver twist or two.
Just googled it: https://en.wikipedia.org/wiki/Attention_Is_All_You_Need
