Josh
@jman4747.bsky.social
He/Him - Watching every* Magical Girl show - Anarchy or bust.
Pinned
Josh
@jman4747.bsky.social
· Sep 3
Good news about the cutting-edge AI tools from the $3-trillion-dollar company.
This is why the bubble is deflating. “AI” boosters will cite this as an example of an LLM that doesn’t run at a loss but then, when you point out the bad answers from cheap models, they cite much more expensive models & usage patterns as being more accurate. They’re stuck with a catch 22.
Reposted by Josh
www.nexusmods.com/morrowind/mo...
I made a mod that replaces cliffracers with Thomas the Tank Engine. I am incapable of learning lessons whenever it involves corporations, because I fundamentally do not view toy company CEOs or media CEOs as people.
#morrowind #thomasthetankengine
I made a mod that replaces cliffracers with Thomas the Tank Engine. I am incapable of learning lessons whenever it involves corporations, because I fundamentally do not view toy company CEOs or media CEOs as people.
#morrowind #thomasthetankengine
November 24, 2025 at 8:52 PM
www.nexusmods.com/morrowind/mo...
I made a mod that replaces cliffracers with Thomas the Tank Engine. I am incapable of learning lessons whenever it involves corporations, because I fundamentally do not view toy company CEOs or media CEOs as people.
#morrowind #thomasthetankengine
I made a mod that replaces cliffracers with Thomas the Tank Engine. I am incapable of learning lessons whenever it involves corporations, because I fundamentally do not view toy company CEOs or media CEOs as people.
#morrowind #thomasthetankengine
Reposted by Josh
"because I fundamentally do not view toy company CEOs or media CEOs as people" is how he finished that sentence
November 28, 2025 at 3:39 AM
"because I fundamentally do not view toy company CEOs or media CEOs as people" is how he finished that sentence
Reposted by Josh
November 27, 2025 at 5:28 AM
Reposted by Josh
Every prompt to chatbot software is “generate me some text.” If it generated you some text, it performed its function correctly. It never lies, it never wants anything, it’s text generating software that generates text.
Not wanting to disappoint you so much that it lies is the last quality I want in a computer.
November 28, 2025 at 3:25 AM
Every prompt to chatbot software is “generate me some text.” If it generated you some text, it performed its function correctly. It never lies, it never wants anything, it’s text generating software that generates text.
Reposted by Josh
ok this time i double checked and the source for this claim is an AI company.
Motivation: we are financially invested in narratives of inevitability
Results: we find AI everywhere
Motivation: we are financially invested in narratives of inevitability
Results: we find AI everywhere
November 28, 2025 at 3:30 AM
ok this time i double checked and the source for this claim is an AI company.
Motivation: we are financially invested in narratives of inevitability
Results: we find AI everywhere
Motivation: we are financially invested in narratives of inevitability
Results: we find AI everywhere
Reposted by Josh
The only reasonable position is that this technology should not be used or deployed under any circumstance by anybody for any purpose, because the makers of this technology are actively working to make any kind of normal assessment of risks and opportunities impossible.
November 27, 2025 at 7:57 PM
The only reasonable position is that this technology should not be used or deployed under any circumstance by anybody for any purpose, because the makers of this technology are actively working to make any kind of normal assessment of risks and opportunities impossible.
Reposted by Josh
This is why being "nuanced" about OpenAI and LLMs is useless. They are engaged in a propaganda campaign to muddy the waters and disinform people. They want you to think "oh, maybe in this special case, oh maybe under these circumstances".
November 27, 2025 at 7:57 PM
This is why being "nuanced" about OpenAI and LLMs is useless. They are engaged in a propaganda campaign to muddy the waters and disinform people. They want you to think "oh, maybe in this special case, oh maybe under these circumstances".
Reposted by Josh
Very clear illustration of how propaganda works – some OpenAI hack publishes a non-peer-reviewed preprint of a paper alongside multiple other OpenAI hacks (most of whom are not clearly listed as working for OpenAI), getting an inexpert journalist to make a deeply unserious claim on social media.
People on BlueSky: AI is useless! A stochastic parrot!
Mathematicians/biologists/physicists: It is already helping us do frontier technical research and in some cases solve open problems arxiv.org/pdf/2511.16072
(There are of course, as always, many caveats, but the paper is genuinely remarkable)
Mathematicians/biologists/physicists: It is already helping us do frontier technical research and in some cases solve open problems arxiv.org/pdf/2511.16072
(There are of course, as always, many caveats, but the paper is genuinely remarkable)
arxiv.org
November 27, 2025 at 7:49 PM
Very clear illustration of how propaganda works – some OpenAI hack publishes a non-peer-reviewed preprint of a paper alongside multiple other OpenAI hacks (most of whom are not clearly listed as working for OpenAI), getting an inexpert journalist to make a deeply unserious claim on social media.
Reposted by Josh
Reposted by Josh
Complete, total, and utter badassery.
November 28, 2025 at 1:20 AM
Complete, total, and utter badassery.
Reposted by Josh
Carbon capture and storage technology is, at its core, a calculated tactic by fossil fuel companies to perpetuate their environmentally destructive practices while giving the illusion that they’re addressing the very crisis they’ve been fueling for over a century.
The false hope of carbon capture and storage
Do we need carbon removal? Almost certainly, but not because we need to offset “hard-to-decarbonize” industries: we need it because we’ve already disrupted the climate to dangerous levels. Just ask an...
canadiandimension.com
November 27, 2025 at 10:10 PM
Carbon capture and storage technology is, at its core, a calculated tactic by fossil fuel companies to perpetuate their environmentally destructive practices while giving the illusion that they’re addressing the very crisis they’ve been fueling for over a century.
Reposted by Josh
I am so torn between wanting to share this with no context so that the reader can have the full experience of developing a relationship with each of the parties in the discussion to this PR, and with knowing that the full impact of what is happening here might evade many without more explanation
DWARF support for macOS and Linux by joelreymont · Pull Request #14369 · ocaml/ocaml
DWARF v5 Debugging Support for OCaml Native Compiler
This PR adds DWARF v5 debug information to the OCaml native compiler, allowing proper source-level debugging in GDB and LLDB.
What's Impleme...
github.com
November 27, 2025 at 7:01 PM
I am so torn between wanting to share this with no context so that the reader can have the full experience of developing a relationship with each of the parties in the discussion to this PR, and with knowing that the full impact of what is happening here might evade many without more explanation
Reposted by Josh
Target could turn around if they stopped selling goods and instead let people pay to watch its executives fend off endless waves of enraged forest creatures.
People would gladly pay $500 a night for a seat to watch the CEO being pursued by dozens of maddened badgers.
People would gladly pay $500 a night for a seat to watch the CEO being pursued by dozens of maddened badgers.
August 20, 2025 at 11:36 AM
Target could turn around if they stopped selling goods and instead let people pay to watch its executives fend off endless waves of enraged forest creatures.
People would gladly pay $500 a night for a seat to watch the CEO being pursued by dozens of maddened badgers.
People would gladly pay $500 a night for a seat to watch the CEO being pursued by dozens of maddened badgers.
Reposted by Josh
"Runctitiononal features"? "Medical fymblal"? "1 Tol Line storee"? This gets worse the longer you look at it. But it's got to be good, because it was published in Nature Scientific Reports last week: www.nature.com/articles/s41... h/t @asa.tsbalans.se
November 27, 2025 at 9:30 AM
"Runctitiononal features"? "Medical fymblal"? "1 Tol Line storee"? This gets worse the longer you look at it. But it's got to be good, because it was published in Nature Scientific Reports last week: www.nature.com/articles/s41... h/t @asa.tsbalans.se
Reposted by Josh
Generative AI is this weird little ferret that is carrying a tall pole with a red flag at the top, and it roots out the worst pre-existing problems in society and plants a little slop-flag in them
eg: how companies have made scientific publishing into a for-profit low quality content mill
eg: how companies have made scientific publishing into a for-profit low quality content mill
"Runctitiononal features"? "Medical fymblal"? "1 Tol Line storee"? This gets worse the longer you look at it. But it's got to be good, because it was published in Nature Scientific Reports last week: www.nature.com/articles/s41... h/t @asa.tsbalans.se
November 27, 2025 at 8:00 PM
Generative AI is this weird little ferret that is carrying a tall pole with a red flag at the top, and it roots out the worst pre-existing problems in society and plants a little slop-flag in them
eg: how companies have made scientific publishing into a for-profit low quality content mill
eg: how companies have made scientific publishing into a for-profit low quality content mill
Reposted by Josh
tired of wealthy people who look like toothpaste models & who's job is 'Doing Nothing' lecturing janitors and bus drivers on how their attitude needs work
November 27, 2025 at 7:58 PM
tired of wealthy people who look like toothpaste models & who's job is 'Doing Nothing' lecturing janitors and bus drivers on how their attitude needs work
Reposted by Josh
Racism aside, the problem with "it's not stolen, it's conquered" is that attitude is nothing but "killing you is an acceptable way to take your stuff" and I'm pretty sure that's bad. "I didn't steal your car, I conquered it" would, uh, not hold up at trial.
November 27, 2025 at 7:34 PM
Racism aside, the problem with "it's not stolen, it's conquered" is that attitude is nothing but "killing you is an acceptable way to take your stuff" and I'm pretty sure that's bad. "I didn't steal your car, I conquered it" would, uh, not hold up at trial.
Seven (7!) of the authors of this paper, which isn't peer reviewed, work for OpenAI... Come on man. That's before we get to which use cases actually used LLM's requiring multi-billion $ datacenter investments and which were "AI" as it's been redefined to mean whatever boosters need it to.
People on BlueSky: AI is useless! A stochastic parrot!
Mathematicians/biologists/physicists: It is already helping us do frontier technical research and in some cases solve open problems arxiv.org/pdf/2511.16072
(There are of course, as always, many caveats, but the paper is genuinely remarkable)
Mathematicians/biologists/physicists: It is already helping us do frontier technical research and in some cases solve open problems arxiv.org/pdf/2511.16072
(There are of course, as always, many caveats, but the paper is genuinely remarkable)
arxiv.org
November 27, 2025 at 7:41 PM
Seven (7!) of the authors of this paper, which isn't peer reviewed, work for OpenAI... Come on man. That's before we get to which use cases actually used LLM's requiring multi-billion $ datacenter investments and which were "AI" as it's been redefined to mean whatever boosters need it to.
Reposted by Josh
It's remarkable because they think advertising lingo like this will get through peer review: "If we had had these interpretations and the recommended next experimental plan from GPT-5 Pro, we would have resolved or hypothesized the mechanistic insights within 19 minutes upon data analysis."
November 27, 2025 at 5:36 PM
It's remarkable because they think advertising lingo like this will get through peer review: "If we had had these interpretations and the recommended next experimental plan from GPT-5 Pro, we would have resolved or hypothesized the mechanistic insights within 19 minutes upon data analysis."
Reposted by Josh
Jeff's replies to this thread are helpful. It's also really important to note that situations in which genAI has been deployed are also often flooding the space with information that still need to be annotated and double checked by human scientists. >>
People on BlueSky: AI is useless! A stochastic parrot!
Mathematicians/biologists/physicists: It is already helping us do frontier technical research and in some cases solve open problems arxiv.org/pdf/2511.16072
(There are of course, as always, many caveats, but the paper is genuinely remarkable)
Mathematicians/biologists/physicists: It is already helping us do frontier technical research and in some cases solve open problems arxiv.org/pdf/2511.16072
(There are of course, as always, many caveats, but the paper is genuinely remarkable)
arxiv.org
November 27, 2025 at 6:35 PM
Jeff's replies to this thread are helpful. It's also really important to note that situations in which genAI has been deployed are also often flooding the space with information that still need to be annotated and double checked by human scientists. >>
Reposted by Josh
The ai bros in my comments are so healthy it’s wild
November 27, 2025 at 6:59 PM
The ai bros in my comments are so healthy it’s wild
Reposted by Josh
They did it again! NYC subway Thanksgiving 2025, via @scootercaster.com!
November 27, 2025 at 4:19 PM
They did it again! NYC subway Thanksgiving 2025, via @scootercaster.com!
Reposted by Josh
He did the same thing to the #StaticCompiler #Julia package developers just a few days later. This shit is unacceptable.
github.com/tshort/Stati...
github.com/tshort/Stati...
Analysis workflow and executable generation by joelreymont · Pull Request #180 · tshort/StaticCompiler.jl
Add comprehensive analysis workflow: quick_check utilities, suggestions/macros,
CLI tooling, binary size analyzer, REPL helpers, and tutorial examples.
Introduce templates/presets for common scenar...
github.com
November 25, 2025 at 2:51 AM
He did the same thing to the #StaticCompiler #Julia package developers just a few days later. This shit is unacceptable.
github.com/tshort/Stati...
github.com/tshort/Stati...
Reposted by Josh
This PR is like the perfect distillation of LLM-assisted hubris. Dev who has never written a compiler or even studied the DWARF spec vibe codes out 13k LOC of "support" in OCaml with a bunch of shallow test cases, and doesn't think he needs to understand his own PR because "Claude understands it!"
One of the many joys of using AI for programming is the creation of huge PRs on complex topics that the authors barely understand, but still suggest "because they work". Here's a great example from #OCaml github.com/ocaml/ocaml/...
Kudos to OCaml's maintainers for handling this so gracefully.
Kudos to OCaml's maintainers for handling this so gracefully.
DWARF support for macOS and Linux by joelreymont · Pull Request #14369 · ocaml/ocaml
DWARF v5 Debugging Support for OCaml Native Compiler
This PR adds DWARF v5 debug information to the OCaml native compiler, allowing proper source-level debugging in GDB and LLDB.
What's Impleme...
github.com
November 24, 2025 at 8:46 PM
This PR is like the perfect distillation of LLM-assisted hubris. Dev who has never written a compiler or even studied the DWARF spec vibe codes out 13k LOC of "support" in OCaml with a bunch of shallow test cases, and doesn't think he needs to understand his own PR because "Claude understands it!"