Tell HN: Reddit JSON endpoints might have been turned off
github.com
github.com
Using a known API key might be a workaround, but it feels more "beginning of the end" rather than a proper workaround.
The workaround that another contributor found (URL-encoding slashes) has been patched. Having a Reddit account will be the only way these frontends will continue working it seems.
I've found myself using LLMs for things I'd normally search Reddit for because I'm tired of being pushed onto New Reddit, and I’m tired of private profiles hiding bots, astroturfers, and other bad actors.
Providing I treat it with some scepticism, why bother visiting Reddit when my LLM uses their data without me having to deal with their anti-patterns?
And if I (someone who prides themselves on trying to make decent contributions) am no longer visiting the site, then I’m no longer contributing either.
Reddit absorbed a large portion of the former webforum audience, and is unusually focused on maintaining human contributions as the primary dataset for visitors, so people get sucked into the convenience trap of not having to host a phpBB or wiki, to solve the discovery problem, and to solve the AI scraper problem.
> my LLM uses their data without me having to deal with their anti-patterns
Do you believe that Reddit is largely failing to prevent LLM usage of current/timely Reddit content, replacing direct app/website usage such as in your case, by a significant (20% or more) portion of their active posting userbase? I would wonder if 20% of their userbase is even able to run their own LLM at all, which leads me to consider that your use case is quite rare — and may explicitly be one Reddit wishes to see evicted.
(If you were referring to people who use LLMs that pay Reddit for access, then Reddit’s already made their money and any cheapening of the site’s value to a hobbyist is irrelevant versus the value of any amount of human-generated content to their LLM customers.)