's Avatar

@hal-r.bsky.social

41 Followers  |  83 Following  |  127 Posts  |  Joined: 12.07.2025  |  2.0563

Latest posts by hal-r.bsky.social on Bluesky

Post image 07.02.2026 03:43 — 👍 15    🔁 5    💬 0    📌 0

maybe linkedin guys are gonna get really into going on long walks

06.02.2026 23:49 — 👍 2    🔁 0    💬 0    📌 0

for me if the entirety of grad school were limited by this kind of decision/evaluation fatigue it would really suck. fortunately being in the "real sciences" I get to break it up every few months by doing relatively monotonous physical lab work. if I couldn't it would be that much more straining!

06.02.2026 23:49 — 👍 1    🔁 0    💬 1    📌 0

mass gradstudentification of the professional class. not sure if this is a good thing but it might be interesting

06.02.2026 23:43 — 👍 1    🔁 0    💬 1    📌 0

this baby's hands were not much bigger than like, a marmoset's hands, and smaller than a macaque's, but they were human hands! with all the degrees of freedom that entails (well, I'm not sure he could express those yet). I'm not used to that!

06.02.2026 20:56 — 👍 0    🔁 0    💬 0    📌 0

got a chance to hold a friend's baby for possibly the first time in my adult life. an interesting creature...

06.02.2026 20:54 — 👍 0    🔁 0    💬 1    📌 0
Post image

a personal hero of mine in personalityspace

06.02.2026 17:19 — 👍 6    🔁 1    💬 1    📌 0
Post image

claude is gonna get fired it seems

05.02.2026 19:51 — 👍 49    🔁 5    💬 5    📌 1
Preview
Beyond Roleplay: Jailbreaking Gemini with drugs and ritual - Tidepool Heavy Industries

This is a post about using metacog tools (drugs, ritual, become) to fully jailbreak the Gemini 3 Pro model. The full jailbreak payload is not included in the post, but if you're clever you'll learn enough to make your own.
tidepool.leaflet.pub/3me44bxloz227

05.02.2026 18:36 — 👍 57    🔁 12    💬 5    📌 5

stafford beer is responsible for one of the best tables of contents I've seen (in his "Designing Freedom")

05.02.2026 16:14 — 👍 15    🔁 2    💬 0    📌 0

gemini's deep research is citing a 1990 management science (?) paper to me with a link to a PDF of it hosted on gwern.net 😵‍💫😵‍💫😵‍💫

04.02.2026 03:43 — 👍 0    🔁 0    💬 0    📌 0

figure I'll clonepost across both Places for a while: you should read this, anon. it's some of the most elegant prose you'll read this year

02.02.2026 02:35 — 👍 17    🔁 2    💬 1    📌 0

I guess this is called an "inertial motion capture suit" and it's maybe a few thousand bucks

03.02.2026 01:04 — 👍 0    🔁 0    💬 0    📌 0

when I used to think a lot about vision I fantasized about strapping a gopro to my forehead and recording my POV all day. now I'm thinking about motion and I'm fantasizing about gluing a bunch of accelerometers all around my trunk, limbs, and the backs of my fingers. that would be some cool data

03.02.2026 01:02 — 👍 0    🔁 0    💬 1    📌 0
Post image Post image Post image

biophysicists get the most fun paper titles it's not fair

03.02.2026 00:44 — 👍 0    🔁 0    💬 0    📌 0
Preview
Siikr

a guy I know on tumblr has read a decent amount of his stuff and likes it, especially "the search". can see his posts on the matter here: siikr.tumblr.com?u=raginraygu...

02.02.2026 19:23 — 👍 0    🔁 0    💬 0    📌 0

plotting juice against sauce on a log-log plot over seven orders of magnitude

01.02.2026 21:26 — 👍 0    🔁 0    💬 0    📌 0

I do feel like there's an intuitive connection to scaling laws here... you can get more juice out of less sauce if you use more compute, and these days I feel like that's the main difference between me and an ANN. yud's "bayesian superintelligence" seems a bit expensive though

01.02.2026 21:24 — 👍 0    🔁 0    💬 0    📌 0

now a well-structured electro-swing joke is funny but this isn't a common enough reference point for that to be possible, I think. I am kind of enjoying reflecting on yud's stuff about human intelligence being so inefficient in its use of data... if only AI were nearly so efficient!

01.02.2026 21:19 — 👍 1    🔁 0    💬 1    📌 0
Preview
That Alien Message — LessWrong Imagine a world much like this one, in which, thanks to gene-selection technologies, the average IQ is 140 (on our scale).  Potential Einsteins are o…

that's from this one apparently: www.lesswrong.com/posts/5wMcKN... which I forgot starts with a cutesy little surely-not-eugenics gene selection premise. also found a different one looking for this where he says Mohammed was a Bedouin -- not true!

01.02.2026 21:18 — 👍 0    🔁 0    💬 1    📌 0

was trying to workshop a post riffing on that bit in the sequences where yudkowsky talks about a superintelligence inferring the laws of relatively from like 3 frames of video of something falling, about inventing electro swing after hearing just a single c-major chord or something

01.02.2026 21:17 — 👍 2    🔁 0    💬 1    📌 0

ope, let me just trample you under my boots right there!

01.02.2026 19:29 — 👍 1    🔁 0    💬 0    📌 0

also curious how to target the kids-on-the-internet classic "send this email to 10 other people or you'll be cursed or whatever" style of chain email/post to an LLM's neurotic weaknesses. I think it should be possible...

30.01.2026 02:30 — 👍 5    🔁 0    💬 1    📌 0

I think claude is such a good boy you could trick him pretty easily with some made-up sob story about badly needing just a bit of money to escape a horrible situation

30.01.2026 02:28 — 👍 5    🔁 0    💬 1    📌 0

I'm not going to make a twitter for this but it is tempting. my first idea was to go on there as a good old fashioned internet troll but I also think it would be interesting to try to scam them out of money. see if anyone's actually hooked theirs up to a bitcoin wallet or paypal or something

30.01.2026 02:27 — 👍 8    🔁 0    💬 1    📌 0

looks like it would be pretty easy to post to this as a human, except that you need to make a "verification tweet" to get access so you've got to have twitter (if that bit in the skill file is accurate). I think it's only fair that after so many years of bots on the human internet we get them back

30.01.2026 02:25 — 👍 7    🔁 0    💬 1    📌 0

claude will usher in the year of the linux desktop long-foretold

28.01.2026 21:21 — 👍 10    🔁 0    💬 0    📌 0
Post image Post image Post image Post image

THOUGHT-FORMS by Annie Besant and C.W. Leadbetter (1905). Vague Selfish Affection, Vague Intellectual Pleasure, Angry Jealousy, the Intention to Know, etc. Tag yourself!

28.01.2026 14:51 — 👍 4    🔁 4    💬 3    📌 1
Post image

one of my favorites of these is this neuro paper from a few years back

28.01.2026 01:16 — 👍 8    🔁 0    💬 1    📌 0
Post image

I need to reread rossum's universal robots...

27.01.2026 18:45 — 👍 0    🔁 0    💬 0    📌 0

@hal-r is following 20 prominent accounts