Content
@
https://obsidian.md
0 reply
0 recast
0 reaction
kepano
@kepano
there's been a sharp acceleration in Obsidian community plugin submissions, in part due to the rise of AI coding tools it's a "good problem to have" but it's overwhelming our review process — our human reviewers can't keep up what should we do? 1. lower the bar for plugin review, introduce more community heuristics 2. invest in AI code-review tools to try and automate the process (not easy) 3. ?
7 replies
35 recasts
66 reactions
chris 🤘🏻
@ckurdziel.eth
Could publish a queue publicly of submissions and let the community upvote/help prioritize certain extensions, then allow others to be installed with the caveat they are not reviewed
1 reply
0 recast
6 reactions
Garrett
@garrett
#2 Fight fire with fire lol
0 reply
0 recast
1 reaction
Thumbs Up
@thumbsup.eth
Three different badges. 1. AI reviewed 2. Community reviewed 3. Obsidian team reviewed Ideally a plugin of any quality eventually has all three. Like Michelin stars but diversified.
1 reply
0 recast
12 reactions
Trigs
@trigs
I like the suggestion of 3 badges from @thumbsup.eth as well as @ckurdziel.eth idea of a 'community queue' or maybe even some kind of bounty board where people can signal how much they like a submission to drive attention from community reviewers. I think there's a really great opportunity for building reputation around this as well, so that community reviewers can be intrinsically motivated to provide good, thorough reviews as well as adding more contextual weight to the aforementioned badges. Been toying around with ideas for this on a few projects. https://warpcast.com/thumbsup.eth/0x9d45b59b
0 reply
0 recast
2 reactions
Adi Seredinschi
@adise
#2
0 reply
0 recast
0 reaction
Adam
@adam-
3. Have applicants self tag their plugins that used AI coding tools. Stipulate that there's no penalty for this, and incentivize them for being honest about it, saying it will help with review process turn around. Subject those submissions to your custom A.I review process. While it's not entirely simple to automate the process, its not as hard either if you can get the right workflow going. Training an LLM on guidelines and then having submissions run through it is doable
0 reply
0 recast
0 reaction
luminoir
@luminoir
Introduce metaanalytics to track uninstalls, make user reviews of plugins more convenient to improve signal to noise ratio
0 reply
0 recast
0 reaction
Murdo
@murdo
Agents make #2 easier, speaking from brief experience.
0 reply
0 recast
0 reaction
