I tried it with exactly one query, something specific that had come up recently in my work. I was looking for one sentence of answer. It was terrible, giving me 500 words of blather, much of which which was irrelevant and some of which was 100% wrong. It was absolutely the arrogant kid who had skipped most of the lectures but who expected to be able to BS through the exam enough to pass the class.
For my needs it was a pure waste of time, and it would have been a bigger waste of time had I not already known enough to judge its output. So I would call this worse than Google and also worse than nothing at all. I suspect this is an inherent problem with LLMs, not something fixable. But in the spirit of constructive criticism, I'd suggest you consider that for programming use cases, no answer is better than a bad answer.
I asked for alternatives to a python library. I did not turn on expert mode because it wasn't clear to me what that meant: expert in the topic, expert in using your tool, maybe something else. I tried turning that on just now and it gave me an answer that looked worse, but so slowly that I gave up before I got to the end.
> Fair as someone coming in blind, but the post here did explicitly tell you to use it and why.
A protip for you: there are few better ways to make a bad product than complaining that the users are doing it wrong. The users are going to keep using it like users do. You either adapt the product, filter for a different set of users, or expect to keep keep generating bad user experiences.
Here, I clicked on the product link on the HN home page, only later going to the discussion that you apparently wanted me to read first. If you really want me to knowt that first, either make it the default or put it on the product page, not buried in 6 paragraphs of gray-on-cream text on a page I may not see until after I've tried it.
I said it was fair as someone coming blind, but you came to a show hn and didn't read the post, had a problem with something you didn't understand and then complained about it. You may find some benefit in reading docs when having problems with tools.
I have nothing to do with phind by the way.
> gray-on-cream text on a page I may not see until after I've tried it.
I'm on board with complaints about hns terrible accessibility.
They asked for user feedback. I gave them user feedback, using it a typical user would. If they believe the right way to use their product is to require reading an HN post first, they are welcome to put that on the homepage. If they don't, then what they said here is irrelevant.
As I said, fair comment for someone coming in blind. But perhaps it would have been more useful if you'd used the feature they were announcing before commenting on the thread about that announcement.
> If they don't, then what they said here is irrelevant.
Then show hns should have no text content at all.
Look, this is quite simple. It's totally fair to explain the confusion about what expert mode means. It's totally fair to say hns ui is absolutely awful.
It's just much less useful if you come to a show hn about a feature launch, don't use the feature being launched then complain about it without providing enough information to replicate the problem.
I won't respond from this point as it's getting rather circular unless you really want me to.
For my needs it was a pure waste of time, and it would have been a bigger waste of time had I not already known enough to judge its output. So I would call this worse than Google and also worse than nothing at all. I suspect this is an inherent problem with LLMs, not something fixable. But in the spirit of constructive criticism, I'd suggest you consider that for programming use cases, no answer is better than a bad answer.