61–70 of 196 posts
Earlier quoted context omitted.
This is the main use I have for browser addins that block sites in search results.
Google and DDG should add a Kagi feature, where you can set the priority of the sites or even block them.
I really need to try using Kagi more.
I got to top 10% about ten years ago, mostly through basic questions and answers that you could easily do when StackOverflow started. I haven’t touched my account in years, and I am now in top 8% with about 5000 reputation, because old questions and answers keep getting upvotes.
I would think few people have much incentive to try and gain StackOverflow reputation, so what keeps StackOverflow going?
Personally the only websites I see displacing SO on Google results are sites that reuse SO content and game Google to get ad revenue.
The only outcome is see from that attitude is it fueled hundreds of spammy SO clones with names like “nerdsolution” and “geekanswer”.
We’d be better iff with SO keeping the data private.
It just doesn’t feel as useful as it did.
And for most programming questions I go to ChatGPT or GitHub search (which is utter garbage but I think a new version is coming).
Yes I've switched to chatgpt for all my programming questions nowadays because of many reasons: 1. It understands the versions more clearly, it won't generate the code for Bootstrap 2 when my question is about Bootstrap 5 2. Asking follow up questions is easy. So the code didn't work. I tell it that and instantly it tells me what I need to do to fix it (leave missing config file, etc) or gives me an alternate. 3. Its…
Chat gpt will never get smarter though (learn anything new), unless actual people use stack overflow to provide the training data. The most important comment on SO is the one where the OP says "I did this and that, and can confirm it's working". There is no such data when you only use the chat bot. It's a major problem. Perhaps the solution is a more tailored "stack gpt", where the "conversations" are published along…
For instance, I needed to write a simple script that gave me all of the roles containing a list of policies where you can specify multiple policies from the command line using “-p” multiple times.
If I look it up on Google, I get this response which is close enough
https://stackoverflow.com/questions/66127551/list-of-all-rol...
And I still have to make slight modifications.
Of course I could write the entire thing myself just by looking at the boto 3 docs.
But instead I told ChatGPT
“ Write a Python script that returns a comma separated list of arns of all AWS roles that contain policies I specify with the “-p” parameter using argparse”
Then I told it, “that won’t work with more than 50 roles”. It then corrected the script and used a paginator.
Yes, I had to know enough to recognize the bug. But I count have written or modified a Python script that fast and I write boto3 based Python scripts all of the time.
I definitely could have told it that the first version wasn’t returning all of the results and it would have corrected itself like it did and added a paginator
Earlier quoted context omitted.
Google and DDG should add a Kagi feature, where you can set the priority of the sites or even block them.
I just started using Kagi, and the ability to just bump Wikipedia as first results has been great so far. I really need to try using Kagi more.
Earlier quoted context omitted.
This is the main use I have for browser addins that block sites in search results.
Not a browser add-on but changing to self-hosted searxNG. Proxy searches to your engines of choice, filter and rewrite urls to your liking (like yt->piped, medium->scripe.rip).
You can setup your own self-hosted instances or existing third party ones. It's a game changer both for privacy and UX.
I think the experts basically got bored and left. I've asked a couple of questions I'd hope would be fairly simple (like how to run a program in cgroups v2), set a big bounty on them, and never got a single useful answer.
Might it be time to introduce financial bounties?
A lot of answers are now out of date. Some dangerously so. And its not just 'fast moving' languages. Bash, nginx, docker, gcp, bluetooth. Much of the landscape of these tools has changed since say 2015, but that is where most of the 'core' answers are from. Many have notes that they are out of date in sub-comments. But its hard to be noticed against a 700 upvote selected top answer.
Out of date answers are useful when you need to fix something based on old / un-updated OS and/or software. Not every installation uses the latest versions of everything.