Bluesky Alt Text Stream
Key topics
The Bluesky Alt Text Stream is a real-time feed of images posted to Bluesky with their corresponding alt text, sparking discussions on content moderation, AI-generated alt text, and the challenges of handling sensitive content.
Snapshot generated from the HN discussion
Discussion Activity
Active discussionFirst comment
12m
Peak period
16
0-2h
Avg / period
3.6
Based on 25 loaded comments
Key moments
- 01Story posted
Sep 25, 2025 at 7:36 PM EDT
3 months ago
Step 01 - 02First comment
Sep 25, 2025 at 7:49 PM EDT
12m after posting
Step 02 - 03Peak activity
16 comments in 0-2h
Hottest window of the conversation
Step 03 - 04Latest activity
Sep 26, 2025 at 12:55 PM EDT
3 months ago
Step 04
Generating AI Summary...
Analyzing up to 500 comments to identify key contributors and discussion patterns
Want the full context?
Jump to the original sources
Read the primary article or dive into the live Hacker News thread when you're ready.
> This demo is part of Image descriptions on Bluesky: not bad, could be better - see the blog post for more details!
I do think it'd be nice if it didn't scroll the log automatically, IF the user has scrolled at all (looking at backlog). A few ways to go about that
But I found that even that was not easy to do with "traditional" OCR, notes here: https://digitalseams.com/blog/image-transcription-humbled-me
Alt text is not there to explain the joke, that would imply cripples are stupid.
AI is fine, if you want to go against the trillions $ come back with a proof.
LLMs fail, wow easy to find, I'll bet alt text on random internet images fails at over 10 times the rate, most are null.
https://www.w3.org/WAI/tutorials/images/decision-tree/
Same with closed captioning!
And then the alt text Nazis can be quiet, too!
Authors can often use an LLM to save time by outputting an initial draft an author can fix.
Readers can also use their own LLM as a supplement, to ask it questions about details in an image that wouldn't be in the alt text. LLMs will get things wrong but are generally good enough for such supplementary use.
> at://did:plc:4zd6mbhgvmjri5rmlsyeq6go/3lzpamhfmfc2d: a mostly naked vivi (she/her) domming a bound matchstick (he/him) in a neon-lit nightclub called PONYBOYS
> at://did:plc:4zd6mbhgvmjri5rmlsyeq6go/3lzpamhfmfc2d: same pic but made absolutely clear that matchstick is into it and consenting :)
I can understand why artists who enjoy the non-con kink dislike this policy, but I think people can also appreciate where this policy comes from. I can also understand why other NSFW artists are anxious that they need to indicate consent, but there is, unfortunately, some amount of judgment that the moderation team has to make on whether something is violative noncon.
Cool website, regardless.
(I think I did this to get more screen real estate or something, but I should probably explain it on the page)
Thanks!
What I think they’re grappling with are:
* Edge cases & slippery slopes: Distinguishing “fictional consent,” “fantasy” vs “non-consensual content,” especially in illustrated or animated work, is hard.
* Automation risk: Letting AI auto-fill alt text or auto-censor content might introduce errors, overreach, or misinterpretation.
* Creator confidence: If policies or enforcement are too strict or unpredictable, content creators (especially in adult / kink / erotic art spaces) may leave or self-censor.
* Accessibility mandate vs burden: Encouraging alt text is good, but requiring it for every image adds friction; enforcing it strictly might deter visual content.
* Policy signaling: How Bluesky frames its sexual content rules sends a message about what kinds of expression are welcome there (and what kinds are policed).
Sounds like avoiding a cross betwedb 4chan and DeviantArt?