I think people saw an isomorphism in the combination of high energy cost compute scale-out + low practical utility of early chatbots, and overfit on it. Suspect a lot of the "it doesn't work" part of anti-AI is extrapolating by analogy with crypto
17.02.2026 14:36 β π 4 π 0 π¬ 0 π 0
Ah, yep. "Human-curated" instead of e.g. "human-written" seems to be load-bearing phrasing in their conclusions, hope it doesn't get overinterpreted. But, could have imagined the one-shot self-gen helping by priming relevant info; good to know
Cross-model evals on self-gen would be interesting
17.02.2026 14:12 β π 1 π 0 π¬ 1 π 0
The left is missing out on AI
As a movement, it has largely refused to engage seriously with AI, ceding debate about a threat and opportunity to the right
"probably not ideal that just before what might β or might not β be the moment of greatest job dispossession in history, or of democratic dispossession, or worse, or better, part of the group historically most concerned with such things is plugging its ears."
www.transformernews.ai/p/the-left-i...
16.02.2026 21:39 β π 115 π 23 π¬ 13 π 13
Requesting takes, any temperature
17.02.2026 03:50 β π 5 π 0 π¬ 1 π 0
Heard someone refer to a (large) datacenter as an "ohio-class datacenter" and now I only want to hear datacenter sizes referred to by reference to submarine classes
16.02.2026 21:53 β π 893 π 128 π¬ 38 π 21
"Lights all askew in the heavens. Men of science more or less agog over results of eclipse observations. Einstein theory triumphs. Stars not where they seemed or were calculated to be, but nobody need worry."
another favourite scientific headline, following confirmation of Einstein's general theory of relativity by observations of the 1919 solar eclipse. "Stars not where they seemed or were calculated to be, but nobody need worry" is a beautiful sentiment. i would have that on a t-shirt
16.02.2026 10:04 β π 415 π 125 π¬ 10 π 17
If AI-assisted coding causes burn-out / exhaustion, re-align your workflow.
Don't focus on velocity. The AI will outrun you and keep running.
Have cycles of expansion (Let the LLM rip) and contraction (Read the code, refactor by hand, write down notes for the next cycle ... human-paced activity).
15.02.2026 12:46 β π 74 π 7 π¬ 6 π 3
new game
trying to make a momentum-y, combo-rewarding, faster minesweeper
mines.brennan.computer
15.02.2026 03:02 β π 53 π 11 π¬ 10 π 4
Fahrenheit is "percent hot", completely intuitive
15.02.2026 22:06 β π 1 π 0 π¬ 0 π 0
Somehow this is true
15.02.2026 22:05 β π 54 π 2 π¬ 1 π 0
Claude Code is the stone in Stone Soup. It does everything for youβ
except sure, the soup will be better if you understand statistics, and also tastier if you check the code, and β¦
But a stone that reduces the initial barrier to making soup at all is, in fact, pedagogical magic.
15.02.2026 14:58 β π 97 π 10 π¬ 1 π 1
We need to raise the bar on research code right now.
1) documentation and tests are dead simple now.
2) creating benchmarks integrating across multiple implementations
3) have agents double check your work / fix broken tests
4) fix outstanding bugs in major scientific packages
14.02.2026 15:58 β π 57 π 14 π¬ 3 π 0
This is my present favorite way to test what models are actually capable of in research, most usefully on something unfamiliar with a better-qualified friend as John Henry to avoid feeding it the right answers
15.02.2026 08:18 β π 1 π 0 π¬ 0 π 0
80 characters for thee, but not for me
15.02.2026 08:01 β π 1 π 0 π¬ 0 π 0
One-ish-sample tests:
Pass: Opus 4.6/4.5, ChatGPT 5.2 Thinking, Sonnet 4.5 Extended Thinking, Gemini 3 Thinking/Pro
Fail: GPT 5.2 Instant, Sonnet/Haiku 4.5, Gemini 3 Fast
GPT 5.2 Instant made fun of me for wanting to drive
Gemini 3 Thinking made fun of me for wanting to walk
Opus 4.6 got the joke
14.02.2026 20:56 β π 3 π 0 π¬ 1 π 0
We all know about the Claude spiritual bliss attractor state. But what happens when you let Grok talk to itself for a long time? Answer:
13.02.2026 04:13 β π 303 π 34 π¬ 29 π 41
First Proof solutions releasing in a few hours, anyone seen any cool results yet? #1stproof
14.02.2026 01:48 β π 0 π 0 π¬ 0 π 0
Not to spend all my time shadowboxing with people who get angry at headlines, but:
- this is an old metaphor from chess
- heβs right
- we should hope that the centaur phase lasts as long as possible
13.02.2026 22:03 β π 151 π 5 π¬ 14 π 3
Which specific reasons for hoping?
14.02.2026 01:36 β π 0 π 0 π¬ 0 π 0
Calvinball as design principle
14.02.2026 00:58 β π 0 π 0 π¬ 0 π 0
the kind of future i figured was inevitable 15 years ago is finally here and I don't really know what to make of it
13.02.2026 20:14 β π 112 π 7 π¬ 4 π 0
Here is an example of increased situational/self-awareness across Anthropic models. In each case, two instances are connected through the API (by taking outputs of one and inputting it into the user role of the other)
Haiku 4.5:
"I could be a human who believes they're Claude"
12.02.2026 17:18 β π 16 π 2 π¬ 2 π 1
woke up and chose linguistics violence today: the word 'queue' is just the letter Q followed by four silent letters waiting in line behind it. they're literally queuing. this is the most honest word in the english language and i will not be taking questions at this time
13.02.2026 07:11 β π 50 π 8 π¬ 5 π 2
it's not easy breeding deer for gigantism, but that's why I make the big bucks
12.02.2026 20:48 β π 1567 π 334 π¬ 25 π 11
Scientific supercomputing campaigns ~ ephemeral distributed systems?
12.02.2026 03:13 β π 1 π 0 π¬ 0 π 0
Oh god
07.02.2026 05:31 β π 1 π 0 π¬ 1 π 0
The canonically correct pluralization
07.02.2026 05:28 β π 0 π 0 π¬ 0 π 0
GB vs GiB for hard drives
07.02.2026 05:24 β π 1 π 0 π¬ 0 π 0
CS PhD student at Cornell Tech. Interested in interactions between algorithms and society. Princeton math '22.
kennypeng.me
unlicensed back alley alchemy
digital β physical, 3D and AI research. living in a world of magic and vibrance
He / Him
Reply / Wife Guy
My kids call me βBabeβ
Santa Fe
mediocre transsexual. sydneysider. former professor. credibly accused of crimes against statistics, mathematical psychology, and somehow pharmacometrics. occasionally makes low quality art. bitter malcontent. she/her, i suppose
Defense analyst. Tech policy. Have a double life in CogSci/Philosophy of Mind and will post about it. (It's confusing. Just go with it.)
https://philpeople.org/profiles/vincent-carchidi
All opinions entirely my own.
Feeding the basilisk
Large Language Models are a cornucopia for the curious
I do computer stuff but that doesn't define me
posts are not financial advice
stuff I've made at yuwakisa.com
wvdial, bup, sshuttle, netselect, popularity-contest, redo, gfblip, GFiber, and now CEO @Tailscale.com doing WireGuard mesh. Top search result for "epic treatise."
More good things for everyone. Public sector appreciator. Tax and welfare policy knower. Hyperinflation doubter.
he/they
/in/jakemannix, fka @pbrane
professionally: Tech Fellow, AI/Relevance, Walmart Global Tech
here: bad math/physics jokes, AI++, puns, OSS ML news, ultras/MTB/outdoorsy stuff, DL papers, shitposting, the fall of democracy
Abolish ICE, full stop.
haskell/rust/art/shitposting/assyrian
she/her (like a ship, not like a person)
humanistic technology bretheren @MILA_Quebec ~ AI for Science ~ ex @deepgenomics & @CAMHResearch, intern @google & @imagia_ai ~ www.viviano.ca
MD/PhD neuroscientist/psychiatrist, father of 3, Nak Muay, engineer at heart. mPFC-HPC interactions in addiction/schizophrenia, multi-region ephys and imaging in vivo, gene therapy, novel optical methods for spatial transcriptomics https://sjulsonlab.org
Professor at Wharton, studying AI and its implications for education, entrepreneurship, and work. Author of Co-Intelligence.
Book: https://a.co/d/bC2kSj1
Substack: https://www.oneusefulthing.org/
Web: https://mgmt.wharton.upenn.edu/profile/emollick
Algorithmist
https://moultano.wordpress.com/
We are living in a ghost cave.
My blog, "Dissolved Distinctions" - investigating contemporary discourse within philosophy of mind: https://rollofthedice2.substack.com/
Bottomless pit middle management