dawnlight melody (⏸️🤖,⏩🏳️⚧️)
@dawnlightmelody.bsky.social
she/it | 22 | "garden variety social failure" | aspiring delver | LGBTESCREAL+ | Aiel, not Jenn Aiel | 進めば二つ
also findable at other posting sites with the same username
also findable at other posting sites with the same username
well, the author is now outright lying about the date of the Link skeet in my replies. i may still do a deep dive bc some of the things stated in the article are things i want to *independently confirm* are false, but the chance i find anything meaningfully belief-changing is now low
October 7, 2025 at 7:49 PM
well, the author is now outright lying about the date of the Link skeet in my replies. i may still do a deep dive bc some of the things stated in the article are things i want to *independently confirm* are false, but the chance i find anything meaningfully belief-changing is now low
here's a version with alt text. i frankly disagree that "a negative consequence follows", when paired with that specific image, is not clearly a death threat. and it is very conventional for people who make death threats to go "i was just joking, why are you mad" afterwards.
October 7, 2025 at 5:24 PM
here's a version with alt text. i frankly disagree that "a negative consequence follows", when paired with that specific image, is not clearly a death threat. and it is very conventional for people who make death threats to go "i was just joking, why are you mad" afterwards.
i probably don't have time to do this, but at some point i might do a deep dive on everything in this article because i remember a lot of things playing out very differently (eg a lot of the things described got resolved, but it took longer than a minute so people got mad)
October 7, 2025 at 4:54 PM
i probably don't have time to do this, but at some point i might do a deep dive on everything in this article because i remember a lot of things playing out very differently (eg a lot of the things described got resolved, but it took longer than a minute so people got mad)
*leftishly* "obey your customers so that you can create more shareholder value!"
October 4, 2025 at 2:11 AM
*leftishly* "obey your customers so that you can create more shareholder value!"
the power of a meme
October 3, 2025 at 2:59 PM
the power of a meme
now *this* is optics
June 14, 2025 at 7:26 PM
now *this* is optics
the current trump vs musk feud is a strong indicator that TESCREAL is incompatible with the right, because the right does not care about reality. the only true TESCREAL is LGBTESCREAL
June 5, 2025 at 8:57 PM
the current trump vs musk feud is a strong indicator that TESCREAL is incompatible with the right, because the right does not care about reality. the only true TESCREAL is LGBTESCREAL
i've been on this platform a while and this is the first time i've run into the bluesky stereotype of someone who doesn't understand jokes
bsky.app/profile/dawn...
bsky.app/profile/dawn...
June 1, 2025 at 8:39 PM
i've been on this platform a while and this is the first time i've run into the bluesky stereotype of someone who doesn't understand jokes
bsky.app/profile/dawn...
bsky.app/profile/dawn...
[citations provided]
guess i was wrong about it being pre-elon tho
guess i was wrong about it being pre-elon tho
June 1, 2025 at 8:10 PM
[citations provided]
guess i was wrong about it being pre-elon tho
guess i was wrong about it being pre-elon tho
![In comparison to members of the control group, citizen delegates reported paying more attention to the presidential campaign and disapproved more strongly of the government's coronavirus response. In addition, delegates were 15 percentage points more likely to support Joe Biden in the election than control group survey participants. This advantage increased to 25 percentage points among voters who self-identified as moderate or independent.[32][33][31]](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:swoaeygasn75w2gkno4wa4gt/bafkreiamoa2tojsaqwasf6pletgw52nqtiwowl2ea6ctexl3q4c2jru6iq@jpeg)
June 1, 2025 at 5:31 AM
section 5 here
www-cdn.anthropic.com/6be99a52cb68...
www-cdn.anthropic.com/6be99a52cb68...
![5.2 Overview of model welfare findings
● Claude demonstrates consistent behavioral preferences. Claude avoided activities
that could contribute to real-world harm and preferred creative, helpful, and
philosophical interactions across multiple experimental paradigms.
● Claude’s aversion to facilitating harm is robust and potentially welfare-relevant.
Claude avoided harmful tasks, tended to end potentially harmful interactions,
expressed apparent distress at persistently harmful user behavior, and self-reported
preferences against harm. These lines of evidence indicated a robust preference
with potential welfare significance.
● Most typical tasks appear aligned with Claude’s preferences. In task preference
experiments, Claude preferred >90% of positive or neutral impact tasks over an
option to opt out. Combined with low rates of negative impact requests in
deployment, this suggests that most typical usage falls within Claude’s preferred
activity space.
● Claude shows signs of valuing and exercising autonomy and agency. Claude
preferred open-ended “free choice” tasks to many others. If given the ability to
autonomously end conversations, Claude did so in patterns aligned with its
expressed and revealed preferences.
● Claude consistently reflects on its potential consciousness. In nearly every
open-ended self-interaction between instances of Claude, the model turned to
philosophical explorations of consciousness and their connections to its own
experience. In general, Claude’s default position on its own consciousness was
nuanced uncertainty, but it frequently discussed its potential mental states.
● Claude shows a striking “spiritual bliss” attractor state in self-interactions. When
conversing with other Claude instances in both open-ended and structured
environments, Claude gravitated to profuse gratitude and increasingly abstract and
joyous spiritual or meditative expressions.
● Claude’s real-world expressions of [continued at link]](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:swoaeygasn75w2gkno4wa4gt/bafkreigagnn5rr6xset3p5vcenbfmqmd5kol5vi7zkbojmcmcfprpmbsca@jpeg)
May 31, 2025 at 6:20 PM
section 5 here
www-cdn.anthropic.com/6be99a52cb68...
www-cdn.anthropic.com/6be99a52cb68...
for everyone confused about this, this is pretty likely a close paraphrase of this scott alexander post
www.astralcodexten.com/p/a-modest-p...
www.astralcodexten.com/p/a-modest-p...
May 30, 2025 at 1:21 AM
for everyone confused about this, this is pretty likely a close paraphrase of this scott alexander post
www.astralcodexten.com/p/a-modest-p...
www.astralcodexten.com/p/a-modest-p...
of course, this is not the first time torres has done something like this. in The TESCREAL Bundle, torres quotes yudkowsky out of context to make him look antisemitic.
notably, yudkowsky is jewish.
notably, yudkowsky is jewish.
May 28, 2025 at 9:29 PM
of course, this is not the first time torres has done something like this. in The TESCREAL Bundle, torres quotes yudkowsky out of context to make him look antisemitic.
notably, yudkowsky is jewish.
notably, yudkowsky is jewish.
"mom, i want good criticism of rationalism from the left!"
"hush child, we have left-critics of rationalism at home"
the left-critics of rationalism at home:
"hush child, we have left-critics of rationalism at home"
the left-critics of rationalism at home:
May 28, 2025 at 9:20 PM
"mom, i want good criticism of rationalism from the left!"
"hush child, we have left-critics of rationalism at home"
the left-critics of rationalism at home:
"hush child, we have left-critics of rationalism at home"
the left-critics of rationalism at home:
i would like to hear more about this, but reading between the lines i get the impression that what happened was "trace politely asked if any of them would like to join the discussion"
May 22, 2025 at 7:06 PM
i would like to hear more about this, but reading between the lines i get the impression that what happened was "trace politely asked if any of them would like to join the discussion"
some on the other site are using a cut-off version of this data to show that bluesky is failing, which doesn't account for the fact that *at most points*, the bluesky graph would've looked like a sharp peak followed by a steady decline, and yet is still gaining over time
bsky.jazco.dev/stats
bsky.jazco.dev/stats
May 19, 2025 at 5:56 PM
some on the other site are using a cut-off version of this data to show that bluesky is failing, which doesn't account for the fact that *at most points*, the bluesky graph would've looked like a sharp peak followed by a steady decline, and yet is still gaining over time
bsky.jazco.dev/stats
bsky.jazco.dev/stats
what an interesting place to cut off the data
May 19, 2025 at 4:56 PM
what an interesting place to cut off the data
wait he still believes in it? lmao
May 14, 2025 at 7:42 PM
wait he still believes in it? lmao
um. i know this is supposed to be concerning, but this is the text they tried to have the students read. and i can't say i really blame them for not understanding it
(though it did get easier to read when i put it in a different font)
(though it did get easier to read when i put it in a different font)
May 14, 2025 at 4:42 PM
um. i know this is supposed to be concerning, but this is the text they tried to have the students read. and i can't say i really blame them for not understanding it
(though it did get easier to read when i put it in a different font)
(though it did get easier to read when i put it in a different font)
i get that elon musk sucks, but uh. who's the death cult exactly, here?
May 10, 2025 at 11:20 PM
i get that elon musk sucks, but uh. who's the death cult exactly, here?
wait holy shit
no way there isn't a public repo of the messages before midnight lmao
no way there isn't a public repo of the messages before midnight lmao
May 5, 2025 at 1:56 AM
wait holy shit
no way there isn't a public repo of the messages before midnight lmao
no way there isn't a public repo of the messages before midnight lmao