Cool, those would be common information. Easier to find than ever
Hell, if you search it it’ll come up multiple times. Hell, the AI summary will include exactly the lines I’m looking for. Hell, I probably don’t even bother, most of the time I’ll just click through in my IDE and read the relevant bit of the source code myself.
If the answer is in the man pages, the problem is so simple I won’t remember there was a problem
I’m still going to need that single forum post where the author of a framework makes an offhand comment that tangentially hints at the cause of the bug I’m hitting my head against
No amount of man pages are going to tell me I need to downgrade a library, or change a configuration value no one bothered to properly document
That was never easy to find, Google was never good at finding tangential hints. And nowadays you can ask a (local, if you want) agent to analyze the whole library and tell you where the library is crashing. Or open a github issue with the framework to have the author, or the community give you the hint. Good luck doing that on sourceforge.
That was never easy to find, Google was never good at finding tangential hints.
It absolutely used to be, it was a huge part of my workflow. You put in the framework/library and one or two related terms, and it would bring up pages of information. If you dug through that, maybe make the terms more general once or twice, you literally would find some random forum post that gives you exactly what you needed. Usually someone else who solved the problem, but occasionally it was just a little detail in something unrelated. Or sometimes there just really was nothing, now it’s all the time
Hell, you used to be able to paste the error directly with the project name and get that specific error through the stack and line numbers. Now, it just sees an error message and abstractes it into “any error message”. After the first few results, it’ll be random stack errors now also from completely different projects
And I have tried to use a local LLM to dig through results, the problem is the information no longer comes up. Between sites closing themselves off to prevent scraping and Google intentionally demolishing their search capabilities, that entire branch of development skill is just gone
It was a stark change, just around 2 years ago I noticed it happening, and it’s only gotten worse
At this point you really need an AI to read through the source if you don’t want to do it yourself
I guess you could contact the developers, but it’s generally faster to figure it out yourself, so I never got into the habit
Easier than the alternative multiple times over… and these were examples of how powerful it used to be. These were esoteric problems that couldn’t be worked around, either new libraries we were using to do something wired or legacy code we had to interface with
Like I mentioned, normally you could just Google the stack trace and find a discussion on your error immediately. AI is a decent replacement for that at least
I’m really not. This was my daily workflow, one day it was working, the next it was not. There were gradual changes before and after as websites locked themselves down to scraping, but there was a specific update to Google’s search, followed by Bing and the sites that repackage it. I switched to using searxng and quant, but neither of them were ever as good
I couldn’t narrow down to the day, but I could probably pin it down to a specific week if I dug through my notes
I just checked, my man pages are still there.
Cool, those would be common information. Easier to find than ever
Hell, if you search it it’ll come up multiple times. Hell, the AI summary will include exactly the lines I’m looking for. Hell, I probably don’t even bother, most of the time I’ll just click through in my IDE and read the relevant bit of the source code myself.
If the answer is in the man pages, the problem is so simple I won’t remember there was a problem
I’m still going to need that single forum post where the author of a framework makes an offhand comment that tangentially hints at the cause of the bug I’m hitting my head against
No amount of man pages are going to tell me I need to downgrade a library, or change a configuration value no one bothered to properly document
That was never easy to find, Google was never good at finding tangential hints. And nowadays you can ask a (local, if you want) agent to analyze the whole library and tell you where the library is crashing. Or open a github issue with the framework to have the author, or the community give you the hint. Good luck doing that on sourceforge.
It absolutely used to be, it was a huge part of my workflow. You put in the framework/library and one or two related terms, and it would bring up pages of information. If you dug through that, maybe make the terms more general once or twice, you literally would find some random forum post that gives you exactly what you needed. Usually someone else who solved the problem, but occasionally it was just a little detail in something unrelated. Or sometimes there just really was nothing, now it’s all the time
Hell, you used to be able to paste the error directly with the project name and get that specific error through the stack and line numbers. Now, it just sees an error message and abstractes it into “any error message”. After the first few results, it’ll be random stack errors now also from completely different projects
And I have tried to use a local LLM to dig through results, the problem is the information no longer comes up. Between sites closing themselves off to prevent scraping and Google intentionally demolishing their search capabilities, that entire branch of development skill is just gone
It was a stark change, just around 2 years ago I noticed it happening, and it’s only gotten worse
At this point you really need an AI to read through the source if you don’t want to do it yourself
I guess you could contact the developers, but it’s generally faster to figure it out yourself, so I never got into the habit
Yeah, digging through that is the opposite of easy…
Easier than the alternative multiple times over… and these were examples of how powerful it used to be. These were esoteric problems that couldn’t be worked around, either new libraries we were using to do something wired or legacy code we had to interface with
Like I mentioned, normally you could just Google the stack trace and find a discussion on your error immediately. AI is a decent replacement for that at least
This was published in 2011: https://xkcd.com/979/
This was the rule, not the exception. I remember spending a whole afternoon with a PHP error where the error message was in Hebrew. It was bad.
You are seeing the past with rose colored glasses.
I’m really not. This was my daily workflow, one day it was working, the next it was not. There were gradual changes before and after as websites locked themselves down to scraping, but there was a specific update to Google’s search, followed by Bing and the sites that repackage it. I switched to using searxng and quant, but neither of them were ever as good
I couldn’t narrow down to the day, but I could probably pin it down to a specific week if I dug through my notes
You said in a previous post you had to “dig through it”. That’s not needed nowadays. It is measurably faster.