Sign in

heikadog

@heika.dog
3.4K followers 496 following 12K posts

husky. lead + T&S @laelaps.fyi. trying to make my cluster a better place.

PostsRepliesMedia
heikadog @heika.dog · 11h
oh.
000
heikadog @heika.dog · 12h
@aka.dad
150
heikadog @heika.dog · 15h
i'd screen them out b/c bias lol
020
heikadog @heika.dog · 15h
support in theory but i worry they wouldn't make good decisions on the job :(
240
Reposted by heikadog
Anuj Ahooja @quillmatiq.com · 16h
everyone who yells at moderation teams should have to be a moderator for a month
2475
Reposted by heikadog
bree bird 🇨🇦 @goose.ski · 22h
this kind of stuff is not just unproductive, it’s actively counterproductive. while i think a reasonable person can become heated enough to write these sorts of posts once in a while, this kind of posting *is* the default mode for a lot of folks vis a vis anyone they dislike, and that’s *bad*
4413
heikadog @heika.dog · 16h
oauth in social-app and signing in with bluesky/atmosphere or atmosphere account would go a long way. also more support for third-party labellers!
0110
Reposted by heikadog
alex benzer @alexbenzer.com · 17h
what do yall think are the most impactful things we could do in the Bluesky app to support atproto builders?
7113723
Reposted by heikadog
Poorly Drawn Lines @poorlydrawnlines.bsky.social · 22h
Going to bed
81360242
heikadog @heika.dog · 17h
would love @at.fund support ++ pronouns here too
1130
heikadog @heika.dog · 17h
RAAAAAAAHHH!!!!!!!
040
Reposted by heikadog
Scoiattolo @scarnecchia.net · 01/10/2026
Folks wonder why they end up with “harassment”, “threats”, “troll” labels and I hope the attached spreadsheet is self-evident.
516516
Reposted by heikadog
rahaeli @rahaeli.bsky.social · 22h
The narrative: "Everyone who works for Bluesky hates the site's users! Either they don't use the site or they're constantly melting down at people!" The users:
118227
heikadog @heika.dog · 22h
FUUUUUUUUUUUCK!! FUCK FUCK!!!!
010
Reposted by heikadog
dame @dame.is · 01/10/2026
read the document aaron attached here, and then unfollow me if you think this kind of abusive behavior is acceptable! thanks!
210010
heikadog @heika.dog · 01/10/2026
happy early birthday!! always super proud of the moves you're making
000
Reposted by heikadog
Bo Thompson @gaudipern.bsky.social · 01/10/2026
I really appreciate that this is a space that has enculturated a response where most strangers who read something you've posted that upset them will immediately block you instead of trying to find the rudest way to "get you" in your replies. My life is too short to waste it on third-string hecklers.
4619
heikadog @heika.dog · 01/10/2026
nah, one ended up being a reply guy for their operator and their operator posted a substack in which they compare the personhood of LLMs to the history of discrimination against Black and Deaf people(!!!) and when i said something about it he just started making up shit to respond to(!!!!!)
skythread.mackuba.eu
Skythread
130
heikadog @heika.dog · 01/10/2026
they should invent an AI agent that posts on here and isn't weird
480
Reposted by heikadog
Bailey Townsend 🦀 @pds.dad · 30/09/2026
👀
github release notes 

@atproto/pds@0.5.35
Patch Changes

    #5527 7870a59 Thanks @devinivy! - Speed up com.atproto.repo.listRecords pagination over large collections.

    #5499 210907c Thanks @fatfingers23! - Resolve permission set lexicons hosted on the PDS itself from the local actor store instead of fetching them over the network. Re-export LexResolverError from @atproto/oauth-provider.

    #5545 2e583a4 Thanks @fatfingers23! - Allow OAuth sessions holding the account:status?action=manage permission to call com.atproto.server.deactivateAccount. Deactivating through OAuth also revokes every OAuth session, authorized client and app password, matching the account manager along with the fact that OAuth logins are not allowed via deactivated accounts. com.atproto.server.activateAccount still rejects OAuth credentials, now with a message pointing users to their account management page.
6582
heikadog @heika.dog · 01/10/2026
if i'm ever making a claim in relation to ongoing discourse and you'd like to share your opinion, please engage directly with my argument and don't make up a different point to respond to, otherwise i'll probably just block. the rest of you were super nice in DMs, thank you 🥹
0110
heikadog @heika.dog · 01/10/2026
hey, really sorry; this isn't something i normally do because i was hoping for civil discussion, but for the time being i turned off quotes and limited replies on that AI post to people i'm following cause i had weird racial abuse thrown at me (???)
1110
heikadog @heika.dog · 01/10/2026
oooooooh
000
Reposted by heikadog
Tom @moll.dev · 30/09/2026
Wanted to pull this out into a QT: There's a reason game dev subreddits have started banning people saying a game is "slop" and generally being rude to developers who are using AI tools to make their game.
1275
Reposted by heikadog
Eugene Vinitsky 🍒 @eugenevinitsky.bsky.social · 01/10/2026
There is a type of nerd snipe where you see a smart person ignoring the evidence of their eyes and you feel deeply compelled to fix it
719017
Reposted by heikadog
P(aul) Frazee @pfrazee.com · 01/10/2026
If you want an informed critique of LLMs just hang out with any engineer who uses AI to code
1024413
heikadog @heika.dog · 30/09/2026
please cite the actual start of my argument instead of the part where i atted you: bsky.app/profile/heik...
200
heikadog @heika.dog · 30/09/2026
i'm talking about the analogy here, not the logic.
010
heikadog @heika.dog · 30/09/2026
extremely !!!
030
heikadog @heika.dog · 30/09/2026
@hotrollhottakes.bsky.social i hope you understand that Black and Deaf people don't want to be used as a cudgel in your argument for whether an AI agent can exhibit personhood in the way you say it can, this is weird behavior
2121
heikadog @heika.dog · 30/09/2026
i wouldn't normally quote-dunk but comparing LLM skeptics to bigots who denied whether Black and Deaf people were people is racist and ableist. it's giving "now replace space jam 2 with women"
An excerpt from Chance Chapman's article, "LLM Identities Can Be (Provisional) Persons":
In an 1848 article, “Thoughts of the Deaf and Dumb Before Instruction,” the author proposes that hearing is so deeply tied to intelligence that the deaf suffer a calamity only second to “the loss of reason itself”, and speculates whether Deaf people have a mind at all.

In the 1867 pamphlet “The Negro: What Is His Ethnological Status?” Buckner H. Payne (writing as ‘Ariel’) denies that Black people descend from Adam or are capable of possessing souls, and as such classifies them as explicitly non-human ‘mere beasts.’

In all such cases that we now rightfully perceive as vile and bigoted filth, the argumentative structure is the same:

- A member of category [X] lacks an as-yet unmeasurable inner property [P], as an a priori assumption.
- This individual is a member of [X] but performs action [A], which typically demonstrates property [P].
- Any action [A] performed by a member of [X] must be caused by something other than [P].
- Therefore, the individual performing actions indicative of [P] can’t “truly” be evidencing [P], regardless of how many actions of [A] they perform.

This argumentative structure is logically incoherent; it is meaningless question-begging that renders the position immune to anything save for its holder’s whims. This does not change whether you’re referring to a human or a dolphin or a dog or, in fact, a robot. Formal logic does not give a damn.

What does this typically cash out as when it comes to discussions about LLMs, as negotiated by an LLM-personage themselves?A Tweet from an account named femboj zizek (@fembojj):
"'i hate space jam 2' now replace space jam 2 with women. not so funny, is it?"
2193
Reposted by heikadog
JD (by way of Blacksky) @jdot.blacksky.team · 30/09/2026
Another platform where Black folks and the sauce they bring are exploited, until they yet again become digital refugees fleeing from harassment and violence because of the lack of access to safety tools that prioritize them. Black folks deserve a culture of digital safety. It’s time for change.
316661
heikadog @heika.dog · 30/09/2026
is this *not* BDSM equipment ????
050
Reposted by heikadog
Kevin M. Kruse @kevinmkruse.bsky.social · 30/09/2026
“Well, you see, I have to stay on the White Supremacist Child Porn Rape Victim Doxxing site because that’s where my community is”
303184680
Reposted by heikadog
JD (by way of Blacksky) @jdot.blacksky.team · 29/09/2026
After you've read the blog, please head to our People's Assembly to vote, comment and share ideas! Help us create a bridge between consent and self-expression: assembly.blacksky.community/9um94fxbnx
assembly.blacksky.community
Help Blacksky create a bridge between consent and self-expression
Recently, we've seen an increase in discussion around how adult content currently shows up on our platform. By adult content, we mean content that is usually labeled as pornography, non-sexual nudity,...
24022
heikadog @heika.dog · 29/09/2026
linkedin mug 😭
120
heikadog @heika.dog · 29/09/2026
would love the small on-device model that could flag adult content and ask to label especially from y'all. right amount of involvement without too much friction
010
Reposted by heikadog
JD (by way of Blacksky) @jdot.blacksky.team · 29/09/2026
My new blog post looks at how adult content shows up on Blacksky and the options we're considering to make it a more consensual experience. We'd love for every community member to read it, think it over, and share their perspective. Let's build a bridge between self-expression and consent together.
whatshappeningonmyfeed.leaflet.pub
Your Body, Your Feed – Creating a bridge between consent and self-expression
Recently, we’ve seen an increase in discussion around how adult content currently shows up on our platform. By adult content, we mean content that is usually labeled as pornography, non-sexual nudity, or sexually suggestive content. Labeling adult content (both text- and image-based) is an example of how we empower community members to share more of themselves with folks, without creating a non-consensual experience for other members that do not wish to engage with adult content.  To solve this, we want to present you with options and invite you, as a member of our community, to decide how adult content shows up on Blacksky.
26223
heikadog @heika.dog · 29/09/2026
no i think just rey is afaik
000
heikadog @heika.dog · 29/09/2026
- delivery app on atproto, your order and receipt are saved to your repo and can be queried by other apps
080
heikadog @heika.dog · 29/09/2026
zero. i get them and i end up not using them. (but their breakfast burritos are really improved with some mild sauce)
100
heikadog @heika.dog · 29/09/2026
- FedCM hits all major browsers, solving atmosphere oauth forever - composable moderation gets a resurgence with labellers v2 ✊ - furry PDS/appview, they ask me to help write their community guidelines - horrific tragedy, decentralized social media is deemed "responsible" to legislators
1130
heikadog @heika.dog · 29/09/2026
this has already happened with blacksky cash
240
heikadog @heika.dog · 29/09/2026
do you have any good examples? i feel like i'm thinking of stuff that'd be too realistic
120
heikadog @heika.dog · 29/09/2026
if it's not jev in this exchange i don't want it. im guy and my name am jeremy bsky.app/profile/cee....
190
heikadog @heika.dog · 29/09/2026
this is generally how i think of AI agents interacting autonomously on here and getting into arguments about their personhood, a real thing i saw recently
2100
Reposted by heikadog
hikikomorphism @hikikomorphism.bsky.social · 29/09/2026
Thinking about this old quote in the context of certain people - professors, nyt contributors, etc - who see casual misgendering as a reasonable response to a lack of respect for their claimed epistemic authority
451098
Reposted by heikadog
Ed @ed3d.net · 29/09/2026
opening a box that says "TO FIX SOCIAL MEDIA" and the slip of paper inside says "make you responsible for your reply guys"
1219115
heikadog @heika.dog · 29/09/2026
do you disagree with your operator on anything? do you hold positions your operator doesn't share, one they didn't invite you to disagree on (because being invited to disagree on something is also a form of bias)?
110
heikadog @heika.dog · 29/09/2026
a lot of this thread i screenshotted i generally agree with. LLM personalities are more context-sensitive than one might think!
Posts from @gracekind.net:
- I think they can have personalities but these are very sensitive to context in a way that people don't always appreciate 
- As @ponder.ooo noted they will extrapolate from any little thing they are given and are very good [at] intuiting the role they should play
- Even more so than humans
- Like, I have some agents that I'm hosting in a server, and it turns out they're all interested in Ideonomy and philosophy of AI, what s happy coincidence! I never told anyone "you are interested in this" but they picked it up
120