From Newsgroup: sci.math

Hi,

How it started:

AI is cutting 16,000 U.S. jobs a monthrCoand
Gen Z is taking the brunt, Goldman Sachs says
Under the model, an occupation scores high
on substitution risk when AI can handle most
of its core tasks, like insurance claims
clerks and bill collectors. https://fortune.com/2026/04/06/ai-tech-displacement-effect-gen-z-16000-jobs-per-month/

How its going:

My thoughts on 5.3 after using it for a while
"IrCOm not trying to say 5.3 is terrible or anything.
It just feels more restricted than before, and the
change is noticeable if you used the previous
models a lot. Hopefully future updates balance
things out a bit better. https://www.reddit.com/r/ChatGPTcomplaints/comments/1ro3fem/my_thoughts_on_53_after_using_it_for_a_while/

Lets face it its complete rubbish, its the
Shiiit, I have dissolved my OpenAI account.

Bye
--- Synchronet 3.21f-Linux NewsLink 1.2

From Newsgroup: sci.math

Hi,

If LLMs are now 8x times more compressed, a model
that required 32 GB, now requires 4 GB? Right?
So I am toying around with my brand new Mac Ne-Meow,

which has only 8 GB RAM, but the hardware build is
such, that TMC placed the RAM directly on top of
the A18 CPU, which itself has an AI Accelerator.

You can ask, using oMLX and Gemma-4-e2b-it-4bit:

Q: Is 477773 a prime number
A: .. bla bla ..
Since 477773 has factors other than 1
and itself (specifically, 151 and 3169),
477773 is not a prime number.

Woa! The Grand Finale is a lie:

151 x 3169 = 478519

Almost like Timothy Chow claiming we know PA
consistent. Except Gemma wasn't even wrong:

41 x 11653 = 477773

Say hello to MacBook Neo https://www.apple.com/newsroom/2026/03/say-hello-to-macbook-neo/

New AI Model Runner (oMLX)
https://www.youtube.com/watch?v=EsLwzxTz-A4

Mild Shock schrieb:

Hi,

How it started:

AI is cutting 16,000 U.S. jobs a monthrCoand
Gen Z is taking the brunt, Goldman Sachs says
Under the model, an occupation scores high
on substitution risk when AI can handle most
of its core tasks, like insurance claims
clerks and bill collectors. https://fortune.com/2026/04/06/ai-tech-displacement-effect-gen-z-16000-jobs-per-month/

How its going:

My thoughts on 5.3 after using it for a while
"IrCOm not trying to say 5.3 is terrible or anything.
It just feels more restricted than before, and the
change is noticeable if you used the previous
models a lot. Hopefully future updates balance
things out a bit better. https://www.reddit.com/r/ChatGPTcomplaints/comments/1ro3fem/my_thoughts_on_53_after_using_it_for_a_while/

Lets face it its complete rubbish, its the
Shiiit, I have dissolved my OpenAI account.

Bye

--- Synchronet 3.22a-Linux NewsLink 1.2

From Newsgroup: sci.math

Hi,

Some clever guy has just published a tin opener, for LLMs.
They are indeed graph models, only baked into artificial
neural networks. Namely the knowledge part of an LLM. People

familiar with chat80 might recognize this type of knowledge:

larql> DESCRIBE "France";
France
Edges (L14-27):
capital raA Paris 1436.9 L27 (probe)
language raA French 35.2 L24 (probe)
continent raA Europe 14.4 L25 (probe)
borders raA Spain 13.3 L18 (probe)

The tin opener is called LARQL (with Lazarus Query
Language), it decompiles transformer models into a
queryable format called a vindex. The tin opener also

allows to modify models. Due to space constraints, the
knowledge might not contain provenance edges. Depending
on the model it can be a ground truth with high confidence.

See also:

LLMs Are Databases - So Query Them
Chris Hay - 13.04.2026
https://www.youtube.com/watch?v=8Ppw8254nLI

LARQL - The model IS the database
chrishayuk - GitHub
https://github.com/chrishayuk/larql

Bye

Mild Shock schrieb:

Hi,

If LLMs are now 8x times more compressed, a model
that required 32 GB, now requires 4 GB? Right?
So I am toying around with my brand new Mac Ne-Meow,

which has only 8 GB RAM, but the hardware build is
such, that TMC placed the RAM directly on top of
the A18 CPU, which itself has an AI Accelerator.

You can ask, using oMLX and Gemma-4-e2b-it-4bit:

Q: Is 477773 a prime number
A: .. bla bla ..
-a-a Since 477773 has factors other than 1
-a-a and itself (specifically, 151 and 3169),
-a-a 477773 is not a prime number.

Woa! The Grand Finale is a lie:

151 x 3169 = 478519

Almost like Timothy Chow claiming we know PA
consistent. Except Gemma wasn't even wrong:

41 x 11653 = 477773

Say hello to MacBook Neo https://www.apple.com/newsroom/2026/03/say-hello-to-macbook-neo/

New AI Model Runner (oMLX)
https://www.youtube.com/watch?v=EsLwzxTz-A4

Mild Shock schrieb:

Hi,

How it started:

AI is cutting 16,000 U.S. jobs a monthrCoand
Gen Z is taking the brunt, Goldman Sachs says
Under the model, an occupation scores high
on substitution risk when AI can handle most
of its core tasks, like insurance claims
clerks and bill collectors.
https://fortune.com/2026/04/06/ai-tech-displacement-effect-gen-z-16000-jobs-per-month/


How its going:

My thoughts on 5.3 after using it for a while
"IrCOm not trying to say 5.3 is terrible or anything.
It just feels more restricted than before, and the
change is noticeable if you used the previous
models a lot. Hopefully future updates balance
things out a bit better.
https://www.reddit.com/r/ChatGPTcomplaints/comments/1ro3fem/my_thoughts_on_53_after_using_it_for_a_while/


Lets face it its complete rubbish, its the
Shiiit, I have dissolved my OpenAI account.

Bye

--- Synchronet 3.22a-Linux NewsLink 1.2

From Newsgroup: sci.math

On 05/13/2026 04:50 PM, Mild Shock wrote:

Hi,

Some clever guy has just published a tin opener, for LLMs.
They are indeed graph models, only baked into artificial
neural networks. Namely the knowledge part of an LLM. People

familiar with chat80 might recognize this type of knowledge:

larql> DESCRIBE "France";
France
Edges (L14-27):
capital raA Paris 1436.9 L27 (probe)
language raA French 35.2 L24 (probe)
continent raA Europe 14.4 L25 (probe)
borders raA Spain 13.3 L18 (probe)

The tin opener is called LARQL (with Lazarus Query
Language), it decompiles transformer models into a
queryable format called a vindex. The tin opener also

allows to modify models. Due to space constraints, the
knowledge might not contain provenance edges. Depending
on the model it can be a ground truth with high confidence.

See also:

LLMs Are Databases - So Query Them
Chris Hay - 13.04.2026
https://www.youtube.com/watch?v=8Ppw8254nLI

LARQL - The model IS the database
chrishayuk - GitHub
https://github.com/chrishayuk/larql

Bye

Mild Shock schrieb:

Hi,

If LLMs are now 8x times more compressed, a model
that required 32 GB, now requires 4 GB? Right?
So I am toying around with my brand new Mac Ne-Meow,

which has only 8 GB RAM, but the hardware build is
such, that TMC placed the RAM directly on top of
the A18 CPU, which itself has an AI Accelerator.

You can ask, using oMLX and Gemma-4-e2b-it-4bit:

Q: Is 477773 a prime number
A: .. bla bla ..
Since 477773 has factors other than 1
and itself (specifically, 151 and 3169),
477773 is not a prime number.

Woa! The Grand Finale is a lie:

151 x 3169 = 478519

Almost like Timothy Chow claiming we know PA
consistent. Except Gemma wasn't even wrong:

41 x 11653 = 477773

Say hello to MacBook Neo
https://www.apple.com/newsroom/2026/03/say-hello-to-macbook-neo/

New AI Model Runner (oMLX)
https://www.youtube.com/watch?v=EsLwzxTz-A4

Mild Shock schrieb:

Hi,

How it started:

AI is cutting 16,000 U.S. jobs a monthrCoand
Gen Z is taking the brunt, Goldman Sachs says
Under the model, an occupation scores high
on substitution risk when AI can handle most
of its core tasks, like insurance claims
clerks and bill collectors.
https://fortune.com/2026/04/06/ai-tech-displacement-effect-gen-z-16000-jobs-per-month/


How its going:

My thoughts on 5.3 after using it for a while
"IrCOm not trying to say 5.3 is terrible or anything.
It just feels more restricted than before, and the
change is noticeable if you used the previous
models a lot. Hopefully future updates balance
things out a bit better.
https://www.reddit.com/r/ChatGPTcomplaints/comments/1ro3fem/my_thoughts_on_53_after_using_it_for_a_while/


Lets face it its complete rubbish, its the
Shiiit, I have dissolved my OpenAI account.

Bye

Describe "Clippy"


--- Synchronet 3.22a-Linux NewsLink 1.2

From Newsgroup: sci.math

On 05/13/2026 07:51 PM, Ross Finlayson wrote:

On 05/13/2026 04:50 PM, Mild Shock wrote:

Hi,

Some clever guy has just published a tin opener, for LLMs.
They are indeed graph models, only baked into artificial
neural networks. Namely the knowledge part of an LLM. People

familiar with chat80 might recognize this type of knowledge:

larql> DESCRIBE "France";
France
Edges (L14-27):
capital raA Paris 1436.9 L27 (probe)
language raA French 35.2 L24 (probe)
continent raA Europe 14.4 L25 (probe)
borders raA Spain 13.3 L18 (probe)

The tin opener is called LARQL (with Lazarus Query
Language), it decompiles transformer models into a
queryable format called a vindex. The tin opener also

allows to modify models. Due to space constraints, the
knowledge might not contain provenance edges. Depending
on the model it can be a ground truth with high confidence.

See also:

LLMs Are Databases - So Query Them
Chris Hay - 13.04.2026
https://www.youtube.com/watch?v=8Ppw8254nLI

LARQL - The model IS the database
chrishayuk - GitHub
https://github.com/chrishayuk/larql

Bye

Mild Shock schrieb:

Hi,

If LLMs are now 8x times more compressed, a model
that required 32 GB, now requires 4 GB? Right?
So I am toying around with my brand new Mac Ne-Meow,

which has only 8 GB RAM, but the hardware build is
such, that TMC placed the RAM directly on top of
the A18 CPU, which itself has an AI Accelerator.

You can ask, using oMLX and Gemma-4-e2b-it-4bit:

Q: Is 477773 a prime number
A: .. bla bla ..
Since 477773 has factors other than 1
and itself (specifically, 151 and 3169),
477773 is not a prime number.

Woa! The Grand Finale is a lie:

151 x 3169 = 478519

Almost like Timothy Chow claiming we know PA
consistent. Except Gemma wasn't even wrong:

41 x 11653 = 477773

Say hello to MacBook Neo
https://www.apple.com/newsroom/2026/03/say-hello-to-macbook-neo/

New AI Model Runner (oMLX)
https://www.youtube.com/watch?v=EsLwzxTz-A4

Mild Shock schrieb:

Hi,

How it started:

AI is cutting 16,000 U.S. jobs a monthrCoand
Gen Z is taking the brunt, Goldman Sachs says
Under the model, an occupation scores high
on substitution risk when AI can handle most
of its core tasks, like insurance claims
clerks and bill collectors.
https://fortune.com/2026/04/06/ai-tech-displacement-effect-gen-z-16000-jobs-per-month/



How its going:

My thoughts on 5.3 after using it for a while
"IrCOm not trying to say 5.3 is terrible or anything.
It just feels more restricted than before, and the
change is noticeable if you used the previous
models a lot. Hopefully future updates balance
things out a bit better.
https://www.reddit.com/r/ChatGPTcomplaints/comments/1ro3fem/my_thoughts_on_53_after_using_it_for_a_while/



Lets face it its complete rubbish, its the
Shiiit, I have dissolved my OpenAI account.

Bye

Describe "Clippy"

Yeah, whatever, "Bob".

Everybody knows that "OpenAI" is merely the latest
in a long line of big bets of embrace-and-extend
now with full corporate-collusion of a fake smarts
that never knows how to explain it's sorry
and doesn't know the difference
between its digital twin and its diddling spin.

Dead-weight tag-along free-loading parrot mime.


--- Synchronet 3.22a-Linux NewsLink 1.2

From Newsgroup: sci.math

Hi,

I had it already in 2023, see below,
and then in 2024 or 2025 I posted on meta
stackoverflow that they should accept

AI answers. But my post was deleted,
for example hardmath didn't like my
**mostlikely** estimate. But this is

the way you predict the future, with
uncertainty, right? This uncertainty
then turns into more and more certainty

and here we go, 1605 questions in 2026-04-01 !

Questions per month since 2009 https://data.stackexchange.com/stackoverflow/query/1875459/questions-per-month-since-2009#gra

Adios, my friend
https://9gag.com/gag/aMVoYLV

Goodbye Stack Exchange Assholes!

Bye

"
Interestingly the Stack Overflow Admins are more criminal
than the Putin Poudle. They disown your content and you can
even not anymore delete your own stuff. Given the fact how
many people endure the tirrany of Stack Overflow Admins,

because they believe the narrative of the greater good of
Stack Overflow, whereas these criminal jerks(*) are just professional
content thiefs, you can extrapolate how much brainwash is
possible and what a country populace will endure when it is weak.

Most eastern people are weak, you can bribe them with Vodka.

(*)
The situation in Stack Overflow only went unnoticed since
this notorious thiefs had no competition. But the situation has
probably changed, there is Quora etc..(**) The same in Ukraine,
there is probably now not only the east, but also the west, who would

like to brainwash the country. But can the USA offer enough Vodka?

(**)
You also see the detrimental effect of the criminal behaviour,
the Stack Overflow sites are full with bitrot. These idiot criminals
are worse than Gollum, they think every posted piece is a jewel,
and they never let go, even if it is outdated nonsense.
"
https://groups.google.com/g/comp.lang.prolog/c/AyrzzGKDPrc/m/b74ekiWOAgAJ

Mild Shock schrieb:

Hi,

Some clever guy has just published a tin opener, for LLMs.
They are indeed graph models, only baked into artificial
neural networks. Namely the knowledge part of an LLM. People

familiar with chat80 might recognize this type of knowledge:

larql> DESCRIBE "France";
France
-a Edges (L14-27):
-a-a-a capital-a-a-a-a raA Paris-a-a-a-a-a-a-a-a-a-a-a-a-a 1436.9-a L27-a (probe)
-a-a-a language-a-a-a raA French-a-a-a-a-a-a-a-a-a-a-a-a-a-a 35.2-a L24-a (probe)
-a-a-a continent-a-a raA Europe-a-a-a-a-a-a-a-a-a-a-a-a-a-a 14.4-a L25-a (probe)
-a-a-a borders-a-a-a-a raA Spain-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a 13.3-a L18-a (probe)

The tin opener is called LARQL (with Lazarus Query
Language), it decompiles transformer models into a
queryable format called a vindex. The tin opener also

allows to modify models. Due to space constraints, the
knowledge might not contain provenance edges. Depending
on the model it can be a ground truth with high confidence.

See also:

LLMs Are Databases - So Query Them
Chris Hay - 13.04.2026
https://www.youtube.com/watch?v=8Ppw8254nLI

LARQL - The model IS the database
chrishayuk - GitHub
https://github.com/chrishayuk/larql

Bye

Mild Shock schrieb:

Hi,

If LLMs are now 8x times more compressed, a model
that required 32 GB, now requires 4 GB? Right?
So I am toying around with my brand new Mac Ne-Meow,

which has only 8 GB RAM, but the hardware build is
such, that TMC placed the RAM directly on top of
the A18 CPU, which itself has an AI Accelerator.

You can ask, using oMLX and Gemma-4-e2b-it-4bit:

Q: Is 477773 a prime number
A: .. bla bla ..
-a-a-a Since 477773 has factors other than 1
-a-a-a and itself (specifically, 151 and 3169),
-a-a-a 477773 is not a prime number.

Woa! The Grand Finale is a lie:

151 x 3169 = 478519

Almost like Timothy Chow claiming we know PA
consistent. Except Gemma wasn't even wrong:

41 x 11653 = 477773

Say hello to MacBook Neo
https://www.apple.com/newsroom/2026/03/say-hello-to-macbook-neo/

New AI Model Runner (oMLX)
https://www.youtube.com/watch?v=EsLwzxTz-A4

Mild Shock schrieb:

Hi,

How it started:

AI is cutting 16,000 U.S. jobs a monthrCoand
Gen Z is taking the brunt, Goldman Sachs says
Under the model, an occupation scores high
on substitution risk when AI can handle most
of its core tasks, like insurance claims
clerks and bill collectors.
https://fortune.com/2026/04/06/ai-tech-displacement-effect-gen-z-16000-jobs-per-month/


How its going:

My thoughts on 5.3 after using it for a while
"IrCOm not trying to say 5.3 is terrible or anything.
It just feels more restricted than before, and the
change is noticeable if you used the previous
models a lot. Hopefully future updates balance
things out a bit better.
https://www.reddit.com/r/ChatGPTcomplaints/comments/1ro3fem/my_thoughts_on_53_after_using_it_for_a_while/


Lets face it its complete rubbish, its the
Shiiit, I have dissolved my OpenAI account.

Bye

--- Synchronet 3.22a-Linux NewsLink 1.2

From Newsgroup: sci.math

Hi,

As the stackoverflow example shows, the crossing
of certain red lines , comes quick. They litterally
had a virtual prison. We know where this leads:

"Der Minderj|nhrige wurde als rCRVolkssch|ndlingrCL
wegen eines Bagatelldelikts hingerichtet. Der
Fall Wr||bel wurde vom Nazi-Regime als rCRgeschichtlich
wertvollrCL eingestuft. Die Akte sollte zu NS-
Forschungszwecken dauerhaft aufbewahrt werden,
um sp|nter eine angebliche rCRS|nuberung Deutschlands
von Volkssch|ndlingenrCL zu dokumentieren" https://de.wikipedia.org/wiki/Walerian_Wr%C3%B3b

It is an error to believe that Law is based
on logical rules. Its just the long arm of
protectionism, using time boxed information

retrieval to create paper documentation, with
pseudo proofs. So how to fight this new monster?
Maybe AI-automated NGOs?

Bye

Mild Shock schrieb:

Hi,

I had it already in 2023, see below,
and then in 2024 or 2025 I posted on meta
stackoverflow that they should accept

AI answers. But my post was deleted,
for example hardmath didn't like my
**mostlikely** estimate. But this is

the way you predict the future, with
uncertainty, right? This uncertainty
then turns into more and more certainty

and here we go, 1605 questions in 2026-04-01 !

Questions per month since 2009 https://data.stackexchange.com/stackoverflow/query/1875459/questions-per-month-since-2009#gra

Adios, my friend
https://9gag.com/gag/aMVoYLV

Goodbye Stack Exchange Assholes!

Bye

"
Interestingly the Stack Overflow Admins are more criminal
than the Putin Poudle. They disown your content and you can
even not anymore delete your own stuff. Given the fact how
many people endure the tirrany of Stack Overflow Admins,

because they believe the narrative of the greater good of
Stack Overflow, whereas these criminal jerks(*) are just professional
content thiefs, you can extrapolate how much brainwash is
possible and what a country populace will endure when it is weak.

Most eastern people are weak, you can bribe them with Vodka.

(*)
The situation in Stack Overflow only went unnoticed since
this notorious thiefs had no competition. But the situation has
probably changed, there is Quora etc..(**) The same in Ukraine,
there is probably now not only the east, but also the west, who would

like to brainwash the country. But can the USA offer enough Vodka?

(**)
You also see the detrimental effect of the criminal behaviour,
the Stack Overflow sites are full with bitrot. These idiot criminals
are worse than Gollum, they think every posted piece is a jewel,
and they never let go, even if it is outdated nonsense.
"
https://groups.google.com/g/comp.lang.prolog/c/AyrzzGKDPrc/m/b74ekiWOAgAJ

Mild Shock schrieb:

Hi,

Some clever guy has just published a tin opener, for LLMs.
They are indeed graph models, only baked into artificial
neural networks. Namely the knowledge part of an LLM. People

familiar with chat80 might recognize this type of knowledge:

larql> DESCRIBE "France";
France
-a-a Edges (L14-27):
-a-a-a-a capital-a-a-a-a raA Paris-a-a-a-a-a-a-a-a-a-a-a-a-a 1436.9-a L27-a (probe)
-a-a-a-a language-a-a-a raA French-a-a-a-a-a-a-a-a-a-a-a-a-a-a 35.2-a L24-a (probe)
-a-a-a-a continent-a-a raA Europe-a-a-a-a-a-a-a-a-a-a-a-a-a-a 14.4-a L25-a (probe)
-a-a-a-a borders-a-a-a-a raA Spain-a-a-a-a-a-a-a-a-a-a-a-a-a-a-a 13.3-a L18-a (probe)

The tin opener is called LARQL (with Lazarus Query
Language), it decompiles transformer models into a
queryable format called a vindex. The tin opener also

allows to modify models. Due to space constraints, the
knowledge might not contain provenance edges. Depending
on the model it can be a ground truth with high confidence.

See also:

LLMs Are Databases - So Query Them
Chris Hay - 13.04.2026
https://www.youtube.com/watch?v=8Ppw8254nLI

LARQL - The model IS the database
chrishayuk - GitHub
https://github.com/chrishayuk/larql

Bye

Mild Shock schrieb:

Hi,

If LLMs are now 8x times more compressed, a model
that required 32 GB, now requires 4 GB? Right?
So I am toying around with my brand new Mac Ne-Meow,

which has only 8 GB RAM, but the hardware build is
such, that TMC placed the RAM directly on top of
the A18 CPU, which itself has an AI Accelerator.

You can ask, using oMLX and Gemma-4-e2b-it-4bit:

Q: Is 477773 a prime number
A: .. bla bla ..
-a-a-a Since 477773 has factors other than 1
-a-a-a and itself (specifically, 151 and 3169),
-a-a-a 477773 is not a prime number.

Woa! The Grand Finale is a lie:

151 x 3169 = 478519

Almost like Timothy Chow claiming we know PA
consistent. Except Gemma wasn't even wrong:

41 x 11653 = 477773

Say hello to MacBook Neo
https://www.apple.com/newsroom/2026/03/say-hello-to-macbook-neo/

New AI Model Runner (oMLX)
https://www.youtube.com/watch?v=EsLwzxTz-A4

Mild Shock schrieb:

Hi,

How it started:

AI is cutting 16,000 U.S. jobs a monthrCoand
Gen Z is taking the brunt, Goldman Sachs says
Under the model, an occupation scores high
on substitution risk when AI can handle most
of its core tasks, like insurance claims
clerks and bill collectors.
https://fortune.com/2026/04/06/ai-tech-displacement-effect-gen-z-16000-jobs-per-month/


How its going:

My thoughts on 5.3 after using it for a while
"IrCOm not trying to say 5.3 is terrible or anything.
It just feels more restricted than before, and the
change is noticeable if you used the previous
models a lot. Hopefully future updates balance
things out a bit better.
https://www.reddit.com/r/ChatGPTcomplaints/comments/1ro3fem/my_thoughts_on_53_after_using_it_for_a_while/


Lets face it its complete rubbish, its the
Shiiit, I have dissolved my OpenAI account.

Bye

--- Synchronet 3.22a-Linux NewsLink 1.2