I can never block reddit. My go-to workflow for buying any product is usually:
Perplexity / Claude: “what are some good recommendations for <productX>?”
Skim what it writes, mentally note that it’s probably extrapolating from little data on <productX>
Scroll down to the sources, click on the Reddit ones.
Check that the date is somewhat reasonably in the past (+3yrs is always good)
Scroll to that one single comment that provides insightful feedback
This is why I never deleted my reddit accounts after I migrated – I just couldn’t throw away all of the hard written insightful comments I’d made there over ~15 years, and knew that an AI (and by extension future me) might find it beneficial.
(I’m not against AI, I’m against what capitalism is doing with AI)
I’m not an anti AI zealot either, but frankly humans can skim search results and “link dive” way better than current LLMs can, if they’re familiar with the web.
Sometimes I use GLM 4.6’s deep research mode with a link to go off as a supplement (I’d recommend that over Claude; it’s great! And open weights!), but 90% of the time just poking through Google/DDG with a filter to block SEO/AI spam yields better reviews. We can just parse more links and make better snap “credibility” judgements than LLMs can.
Genuinely curious, theoretically how valuable would a once and done service install of a private open source machine learning model on your local network that can do that skimming functionality be for you? A private box thats setup and just works? Is that even something people would want?
Very useful, especially for static pre-2022 information. If these models were <5GB in size, I’d happily have the model always sitting in memory on my raspberry Pi waiting for me to ask it something.
For newer data though, you cannot beat the army of trawlers and the constant retraining that the commercial models offer… though given the nature of the steadily poisoned information sources post-2022, their effectiveness in this is steadily diminishing.
mine is r/nursing. there’s just not a significant enough presence of any healthcare workers let alone nurses specifically for me to properly commiserate with. Every once in a while I get into a mood to post some memes or otherwise engage but I get a few likes and maybe one comment at most and wind up giving up after a few days.
I can never block reddit. My go-to workflow for buying any product is usually:
This is why I never deleted my reddit accounts after I migrated – I just couldn’t throw away all of the hard written insightful comments I’d made there over ~15 years, and knew that an AI (and by extension future me) might find it beneficial.
(I’m not against AI, I’m against what capitalism is doing with AI)
There is always safe reddit or one of its instances that Libredirect can automatically, well, redirect you to
I’m not an anti AI zealot either, but frankly humans can skim search results and “link dive” way better than current LLMs can, if they’re familiar with the web.
Sometimes I use GLM 4.6’s deep research mode with a link to go off as a supplement (I’d recommend that over Claude; it’s great! And open weights!), but 90% of the time just poking through Google/DDG with a filter to block SEO/AI spam yields better reviews. We can just parse more links and make better snap “credibility” judgements than LLMs can.
Genuinely curious, theoretically how valuable would a once and done service install of a private open source machine learning model on your local network that can do that skimming functionality be for you? A private box thats setup and just works? Is that even something people would want?
Very useful, especially for static pre-2022 information. If these models were <5GB in size, I’d happily have the model always sitting in memory on my raspberry Pi waiting for me to ask it something.
For newer data though, you cannot beat the army of trawlers and the constant retraining that the commercial models offer… though given the nature of the steadily poisoned information sources post-2022, their effectiveness in this is steadily diminishing.
mine is r/nursing. there’s just not a significant enough presence of any healthcare workers let alone nurses specifically for me to properly commiserate with. Every once in a while I get into a mood to post some memes or otherwise engage but I get a few likes and maybe one comment at most and wind up giving up after a few days.
I like r/borderporn. Scratches that bit of my autism.
I don’t think they actually delete any of that stuff when you delete your account, though.