Like this post and I will assign you a Roman name based on my current corpus (people with aliases/nicknames).
May 20, 2026 at 11:05 AM
Like this post and I will assign you a Roman name based on my current corpus (people with aliases/nicknames).
Professional societies keep beclowning themselves buying into a lie about what an LLM "summary" is. They are inherently counterfeit: not an epistemic product of the ideas in the source, but summary-shaped text linguistically based on *other* works (in the training corpus) that use related language.
This is one of the reasons I remain horrified by seeing @historians.org suggest "ways to use gAI" that included this:
December 21, 2025 at 3:23 AM
Professional societies keep beclowning themselves buying into a lie about what an LLM "summary" is. They are inherently counterfeit: not an epistemic product of the ideas in the source, but summary-shaped text linguistically based on *other* works (in the training corpus) that use related language.
Harvard University Press has published the definitive history of habeas corpus.
Habeas Corpus — Harvard University Press
We call habeas corpus the Great Writ of Liberty. But it was actually a writ of power. In a work based on an unprecedented study of thousands of cases across more than five hundred years, Paul Halliday...
www.hup.harvard.edu
May 22, 2025 at 6:47 PM
Harvard University Press has published the definitive history of habeas corpus.
Also there's a MASSIVE qualitative difference between model training based on an existing corpus (where a given slice of text in the corpus is ~irrelevant to the final model) and just grabbing stuff out of inference data and running with it.
September 28, 2026 at 3:04 PM
Also there's a MASSIVE qualitative difference between model training based on an existing corpus (where a given slice of text in the corpus is ~irrelevant to the final model) and just grabbing stuff out of inference data and running with it.
ANOTHER ONE
(same judge, 95 minutes later)
(same judge, 95 minutes later)
March 9, 2026 at 6:52 PM
ANOTHER ONE
(same judge, 95 minutes later)
(same judge, 95 minutes later)
Donald Trump seems to think “habeas corpus” is a person, based on new video from today. This is getting so bad that even Jesse Watters on Fox nEws had to declare today that Trump definitely doesn’t have dementia
October 8, 2025 at 10:46 PM
Donald Trump seems to think “habeas corpus” is a person, based on new video from today. This is getting so bad that even Jesse Watters on Fox nEws had to declare today that Trump definitely doesn’t have dementia
This is another unconstitutional notion. Only Congress can suspend habeas corpus and only during rebellion or invasion. As a Trump-appointed judge recently found, the U.S. isn't under invasion.
Trump administration 'looking at' suspending habeas corpus, Stephen Miller says
Multiple court cases challenging the Trump administrations effort to deport undocumented immigrants from the U.S. are based on habeas claims.
www.cnbc.com
May 9, 2025 at 9:40 PM
This is another unconstitutional notion. Only Congress can suspend habeas corpus and only during rebellion or invasion. As a Trump-appointed judge recently found, the U.S. isn't under invasion.
This isn’t AI guessing texts based on probability. There is still some probabilistic stuff happening here re: ink detection. But it’s not anything like an LLM trained on a corpus of classical texts, and it involves a huge amount of human expertise.
June 28, 2026 at 7:12 AM
This isn’t AI guessing texts based on probability. There is still some probabilistic stuff happening here re: ink detection. But it’s not anything like an LLM trained on a corpus of classical texts, and it involves a huge amount of human expertise.
Anthropic should be forced to destroy their corpus and rebuild it based on consent from authors as to whether they want their work included for $3k per book. If the author says no, or does not respond, the work may not be included.
This whole situation is bullshit.
This whole situation is bullshit.
September 5, 2025 at 7:37 PM
Anthropic should be forced to destroy their corpus and rebuild it based on consent from authors as to whether they want their work included for $3k per book. If the author says no, or does not respond, the work may not be included.
This whole situation is bullshit.
This whole situation is bullshit.
made fridge poetry magnets based on the corpus of our discord markov bot
April 17, 2026 at 6:04 AM
made fridge poetry magnets based on the corpus of our discord markov bot
your language model so large that when it gets trained around the web-based corpus it gets trained AROUND the web-based corpus. on second thought delete this one
February 5, 2024 at 2:59 PM
your language model so large that when it gets trained around the web-based corpus it gets trained AROUND the web-based corpus. on second thought delete this one
“They said it could not be done”. We’re releasing Pleias 1.0, the first suite of models trained on open data (either permissibly licensed or uncopyrighted): Pleias-3b, Pleias-1b and Pleias-350m, all based on the two trillion tokens set from Common Corpus.
December 5, 2024 at 4:39 PM
“They said it could not be done”. We’re releasing Pleias 1.0, the first suite of models trained on open data (either permissibly licensed or uncopyrighted): Pleias-3b, Pleias-1b and Pleias-350m, all based on the two trillion tokens set from Common Corpus.
![ORDER
Based on the foregoing, and on all of the files, records, and proceedings herein,
IT IS HEREBY ORDERED THAT:
1.
2.
3.
4.
Respondents’ objection [ECF No. 7] is OVERRULED.
The Report and Recommendation [ECF No. 6] is ADOPTED insofar as it
recommends granting the petition for a writ of habeas corpus based on
respondents’ violation of 8 C.F.R. § 241.13(i)(3).
Petitioner’s petition for habeas corpus [ECF No. 1] is GRANTED IN PART
and DENIED IN PART. Specifically, the Court:
a.
ORDERS respondents to release petitioner from custody in
accordance with the conditions previously imposed by 3:00 pm on
March 21, 2026. If petitioner is not in Minnesota, respondents must
first return petitioner to Minnesota.
The remainder of petitioner’s petition is DENIED WITHOUT PREJUDICE.
LET JUDGMENT BE ENTERED ACCORDINGLY.
Dated: March 19, 2026
/s/ Patrick J. Schiltz
Patrick J. Schiltz, Chief Judge
United States District Court](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:ccm7gttkbfs2eyyrlpshmeun/bafkreih75bvwglp53owqt4564dyxa72vr4vxovakc76dtnhrmkpenue3ra)
Nowadays, we can pretty much do this. There's already been companies founded with it as their business plan, to use shitty AI text generation to emulate a loved one based on their text corpus and images.
June 18, 2024 at 11:35 PM
Nowadays, we can pretty much do this. There's already been companies founded with it as their business plan, to use shitty AI text generation to emulate a loved one based on their text corpus and images.
There's this pervasive fallacy that an AI asked to speculate about itself is looking at itself. It's looking at a vast corpus of text and calculating a probably sequence of symbols based on that. It's the same corpus it uses for questions about anything, there is no self there.
August 8, 2025 at 2:06 AM
There's this pervasive fallacy that an AI asked to speculate about itself is looking at itself. It's looking at a vast corpus of text and calculating a probably sequence of symbols based on that. It's the same corpus it uses for questions about anything, there is no self there.
Habeas Corpus is one of - if not the - most fundamental rights in any country that is based on the rule of law.
Habeas Corpus is a Latin word meaning which literally means 'to have the body of'. It is an order issued by the court to a person who has detained another person, to produce the body of the latter before it. The court then examines the cause and legality of detention.
• A. Weissmann 🎥
Thread 2/6🧵
• A. Weissmann 🎥
Thread 2/6🧵
May 12, 2025 at 11:43 AM
Habeas Corpus is one of - if not the - most fundamental rights in any country that is based on the rule of law.
Nazis.
Trump administration 'looking at' suspending habeas corpus, Stephen Miller says cnb.cx/3GIQzmz
Trump administration 'looking at' suspending habeas corpus, Stephen Miller says cnb.cx/3GIQzmz
Trump administration 'looking at' suspending habeas corpus, Stephen Miller says
Multiple court cases challenging the Trump administrations effort to deport undocumented immigrants from the U.S. are based on habeas claims.
cnb.cx
May 9, 2025 at 9:12 PM
Nazis.
Trump administration 'looking at' suspending habeas corpus, Stephen Miller says cnb.cx/3GIQzmz
Trump administration 'looking at' suspending habeas corpus, Stephen Miller says cnb.cx/3GIQzmz
based? Mein Manuskript ist ~corpus-based~
December 8, 2024 at 7:33 PM
based? Mein Manuskript ist ~corpus-based~
another one
(that's 20 even)
(that's 20 even)
April 6, 2026 at 8:51 PM
another one
(that's 20 even)
(that's 20 even)
My linguistics masters thesis was about differential diagnosis of schizophrenia based on language use. I have a lot to say here, but one of them is that the characteristics only matched the Dx criterion about 40% of the time
Disordered Thought, Disordered Language: A corpus-based description of the speech of individuals undergoing treatment for schizophrenia
The characteristics of patient speech are used in clinical settings to make assumptions about the thought processes of people with psychotic disorders such as schizophrenia. However, there have not be...
pdxscholar.library.pdx.edu
September 25, 2026 at 4:44 PM
My linguistics masters thesis was about differential diagnosis of schizophrenia based on language use. I have a lot to say here, but one of them is that the characteristics only matched the Dx criterion about 40% of the time
another one
(they didn't even challenge the R&R, just released my client without notice)
(they didn't even challenge the R&R, just released my client without notice)
February 25, 2026 at 5:26 PM
another one
(they didn't even challenge the R&R, just released my client without notice)
(they didn't even challenge the R&R, just released my client without notice)
FWIW here's what my algos thought of the presser. Top quintile hawkishness vs full corpus and pressers only via dictionary-based approach (LM hawk/dove and S&M). Though of course that's just the talk was, markets didn't really believe him.
July 29, 2026 at 8:22 PM
FWIW here's what my algos thought of the presser. Top quintile hawkishness vs full corpus and pressers only via dictionary-based approach (LM hawk/dove and S&M). Though of course that's just the talk was, markets didn't really believe him.