#dev 2023-03-20

2023-03-20 UTC
#
@geekplux_cn 在 blog 中尝试用上了 http://Webmention.io，效果不错啊！ 大家 tweet 我 blog 中任意一篇文章的 url，然后 reply, like or retweet 应该都会出现在对应文章最下面的评论区。比如我上一条 tweet。 (twitter.com/_/status/1637630890242187267)
#
@geekplux Finally I’m using http://Webmention.io on my Blog, looks like emebed the whole Twitter as a comment system. (twitter.com/_/status/1637631465994305537)
[chrisbergr] and Soni joined the channel
#
[snarfed] ^ if [aaronpk] quietly added backfeed to wm.io, I am here for it 😎
Soni joined the channel
#
[tantek] I'm guessing that's because they're using wm io for the embed / display, and maybe forgot they setup Bridgy to handle the backfeed
#
[tantek] from Twitter
#
[snarfed] yup, it was a joke
bterry, mro, IWSlackGateway, lagash, geoffo and [manton] joined the channel
#
[manton] Curious if [aaronpk] or anyone has feedback about this… After talking to another developer who wasn’t familiar with IndieAuth, I decided to write a help page from the perspective of thinking in OAuth when using Micro.blog. I was thinking maybe it’s a stepping stone… First use OAuth, then add the “prompt for domain name” at the beginning of the flow so it works with other blogs. https://help.micro.blog/t/oauth-in-micro-blo
mro and [jacky] joined the channel
#
[jacky] This looks good! I think once more people realize that it's _built on something existing_ (which first sometimes requires pointing out what "Sign in with X" might be), then it'd be second nature
mro joined the channel
#
[manton] Cool, thanks.
#
[manton] Grabbing some HTML and looking for <link> tags is easy, but I wonder if it feels kind of foreign to some programmers who are used to platform APIs. Not sure there’s a solution for that other than more docs.
mro, [schmarty], [KevinMarks] and [tantek] joined the channel
#
[tantek] [manton] we have some documentation / libraries for that on the wiki I believe. pretty sure I split that out "some HTML and looking for <link> tags" into its own page
#
[tantek] what are link rels
#
Loqi It looks like we don't have a page for "link rels" yet. Would you like to create it? (Or just say "link rels is ____", a sentence describing the term)
#
[tantek] [manton] we do have https://indieweb.org/link_rel_parser however that assumes you already know that is what you want and you’re in the process of implementing it
#
[tantek] maybe we need something like "link rel discovery" that provides that kind of a focused explanation (why & how), more specific than https://indieweb.org/discovery-algorithms — [manton] would such a page ("link rel discovery") help the kinds of developers you are thinking of?
#
[tantek] what is follow your nose
#
Loqi 👃 follow your nose is an intentional local & web-centric methodology for designing discovery algorithms that start with a specific URL that to discover things about, inspecting its contents, often for specific hyperlink rel values to predictably discover information about & resources for that URL, in contrast to the “well-known” approach of instead checking outside that URL, like using only its domain with a hardcoded path https://indieweb.org/follow_your_nose
#
[tantek] also related (so to speak) ^ [and needs dfn copy-editing to make it fit]
[aciccarello] joined the channel
#
[manton] Thanks, I’ll look through some of those wiki pages, I don’t think I had seen all of them. That /link_rel_parser looks like a great start as a simple intro page.
#
[manton] And to be clear, I’m not sure this is actually a problem. Just was thinking that link rel parsing is a fundamental think that other building blocks sit on.
#
[manton] (fundamental “thing” too)
#
IWDiscordRelay <c​apjamesg#4492> Is this link discovery?
#
IWDiscordRelay <c​apjamesg#4492> https://twitter.com/natfriedman/status/1575631194032549888?lang=en
#
@natfriedman I’ve been playing with using GPT-3 to control a browser the last couple days. Here’s a quick demo. As you can see it's pretty neat! But also quite flakey. Will publish the source code shortly for others to try and improve. https://twitter.com/natfriedman/status/1575631194032549888/video/1 (twitter.com/_/status/1575631194032549888)
#
IWDiscordRelay <c​apjamesg#4492> Watch the associated video.
s[_] joined the channel
#
capjamesg This is cool: https://twitter.com/JosephArhar/status/1567279577160126464?cxt=HHwWgMDUxZeejMArAAAA
#
@JosephArhar I used chrome's hidden=until-found feature in wikipedia's website, so now you can use find-in-page in collapsed mobile sections! https://pbs.twimg.com/tweet_video_thumb/FcAYSMUaIAAI_YO.jpg (twitter.com/_/status/1567279577160126464)
mro and IWSlackGateway joined the channel
#
capjamesg https://news.ycombinator.com/item?id=35237492
#
Loqi [preview] [temp_account_32] It's funny to see how half of HN frontpage in the past weeks has been Show HN thin wrappers around an API call to ChatGPT disguised as full startups. All while HN posters were warning us about a flood of low quality AI generated content, well, here i...
#
superkuh Now we can move on to thin wrappers around llamas and alpacas running locally. Progress.
mro_ joined the channel
#
IWDiscordRelay <c​apjamesg#4492> 😅
[tantek] joined the channel
#
[tantek] capjamesg re: link discovery, not from the summary/description, and not going to watch a video to answer a question. if there's a text transcript you can point to that would make it easier to evaluate.
#
capjamesg [tantek] There is no audio.
#
capjamesg It's a video of a program that uses GPT 3 to control a web browser and perform various actions.
#
capjamesg It can buy things online.
#
capjamesg It was built by the former CEO at GitHub, who is now involved quite a bit in the AI industry I believe.
#
capjamesg My comment was sarcastic. It works by scraping links from a page then using a Python library to follow them :P
#
[tantek] The phrase "GPT 3 to control a web browser" should be a giant red flag
[schmarty] joined the channel
#
prologic Isn't the whole *GPT* and AI all just hype anyway? I mean they're just LLMs running on some pretty beefy hardware with huge amounts of data trained on
[snarfed] joined the channel
#
[snarfed] #indieweb-chat?
#
prologic yeah probably