iknownthing
iknownthing on Hacker News
in my experience, people are not super interested in using LLMs to help them learn beyond the basic chatgpt chat experience.
Posted by
iknownthing on Tuesday June 23, 2026 @03:10PM
from the ask-hn dept.
When LLMs first hit the scene I had assumed one of the big use cases would be to use the LLM's vast knowledge to teach people about subjects they are interested in. I thought there would be some hot app that let you suggest a subject (e.g. CUDA programming) and the app would interactively teach you and test your knowledge along the way. How come this use case never took off?
I've never found it to make a difference
Well that doesn't look like a Ferrari
That page looks vaguely AI generated
Posted by
iknownthing on Monday July 07, 2025 @05:09PM
from the ask-hn dept.
Whenever I see people doing prompt engineering they start with some kind of evaluation dataset, then they refine their prompt to perform well on that evaluation dataset. But isn't this just like training on a test dataset i.e. overfitting?
Posted by
iknownthing on Wednesday June 25, 2025 @09:16PM
from the ask-hn dept.
I wrote a prompt that did batch classification - the prompt contained instructions on how to classify text and 10 input examples for it to classify and it was to return a json string displaying the classifications. It kind of worked by then I realized the individual classification of an individual input was significantly affected by which other 9 inputs it was in the prompt with. In other...
Posted by
iknownthing on Friday April 18, 2025 @01:32PM
from the show-hn dept.
https://hinterviewgpt.com
Hello HN!I've been an active Leetcode user for a long time and while I think it's a great platform to test your skills, I've never really considered it a good resource for learning. Whenever I'd get stuck on a question I'd just look up the answer then try to understand it via reverse-engineering which never seemed very efficient to me. That's why I built HinterviewGPT.HinterviewGPT started as...
Seems like more of a special case than a different thing altogether
I'm curious, does it get triggered when a PR is opened or when it is merged? Because if it is when it is opened, updates to the PR could still get made which I assume would cause updates to the doc changes. Also, what if 2 PRs are opened at the same time? What if a PR is opened but never merged?
Cronenberg also turned down an offer to direct Top Gun believe it or not
If you put it on your resume the right people will probably read it
Lynch had the cojones to show a guild navigator
I worked at a company that created fake job postings for H1B reasons.
Probably better to do it on anonymous social media
Is this not the same problem everyone faces when they retire?
Well they are supposed to be confessions yet he does not confess the worst part of what he witnessed. IIRC he literally says "I'll tell you everything" then proceeds to leave things out. Somewhat contradictory.
I get that but it doesn't even make sense in terms of the dream logic. Bill basically has two "confession" scenes - one with Zeigler and one with his wife and seemingly doesn't mention the stuff at the costume shop in either. However he was open to confess to Zeigler that he knew about the dead woman. This seems contradictory. I think there are two possibilities 1) it was unfinished or 2) it was intentionally left out. Both of which are very interesting.
What I never understood about this movie is how it never connected the pieces. The beginning of the movie when bill was with the drunk women and they say "where the rainbow ends..." it clearly connects to the "rainbow" costume shop later where the sinister stuff with the owners child happens. Then it's learned that the women at the beginning of the movie were the same women who were at the secret society party this clearly connects the secret society to the sinister stuff at the costume shop. So the connections are clear and bill is privy to all of it yet it is never explicitly stated at the end of the movie. Perhaps Kubrick didn't actually finish it.
Posted by
iknownthing on Saturday December 14, 2024 @05:50PM
from the show-hn dept.
https://github.com/alexminnaar/Diagify
Hello HN,I've created Diagify, a CLI for converting natural language to technical diagrams. There are a bunch of diagram-as-code tools out there and I had the idea that it should be possible to use LLMs to generate the diagram code based on a natural language description, execute it, and return the generated image. This is essentially what Diagify does.More specifically, Diagify generates...
People generally use leetcode to prepare for interviews and learning is kind of a side-effect so I'm curious are these questions similar to those you would find in interviews (i.e. like leetcode) or is it more for general learning purposes?
> with layers upon layers upon layers of meaning
What were the layers of meaning? To me, the shining is kubricks most straight-forward movie.
Wish this existed when I was in school
Posted by
iknownthing on Wednesday November 20, 2024 @01:10PM
from the ask-hn dept.
When you ask some of the popular LLMs to generate code it seems like by default it returns something like "Sure!, here is the code below python```<some_python_code>```" but what if you only want the <some_python_code> part? I managed to make it work by including some example outputs in the prompt but I was wondering if there was a best practice for this?
I mostly use it in a dictation sort of way. I tell it what I want, it gives me its first guess. I read it and tell it what to change. I continue this process until it's what I want. Usually faster than writing it myself.
In my experience, most things in large companies are social constructs. Good things are good because they were decided to be good, bad things are bad because they were decided to be bad. Very rarely are they driven by actual metrics.
Maybe the person reading the application was not aware of the competition either.