This cat and mouse game redlib is playing with Reddit is funny

so Reddit finally locked down their JSON endpoints and it was discovered that percent encoding / as %2F worked around it

great, that worked for a while

then they finally caught on to that too. it looked like nothing was really possible to revive the ability to browse reddit without being tracked

so someone threw claude at it and they found the shreddit service exposes the data as raw html, but it's not in bulk so you have to make a bunch of requests to do it @ an endpoint like

/svc/shreddit/comments/r/{sub}/t3_{id}[/t1_{comment}]?sort=...

so they said fuck it, we'll just do that and reddit can deal with the increased load on their servers for being dicks, but it's missing some important metadata and stuff isn't sorted nicely

surprisingly it works without much latency at all because they batch requests

but now they've also found a graqphql endpoint they can use:

> GraphQL comments: Post pages fetch metadata through /by_id and comments through Reddit's Android PostComments operation, using the existing anonymous OAuth client. This handles Reddit's HTTP 403 block on the comments JSON endpoint. Comment sorting, focused threads, and cursor pagination work without a Reddit account or additional configuration. Each request loads up to 200 comments at 10 levels of depth; links load further comments and replies. The converter and rich text renderer are adapted from [polak14's implementation]