Pinned Tweet
What is High-Protein Content (HPC)? It's a premium, high-density, high-quality data asset optimized for training and improving modern large language models. Imagine the value for library collections! #AI #DataForAI #Libraries
I learned about it from @BrianRoemmele
I want to task one of my Grok Bots (Roberto the Record Ranger, reporting to Gregorio) to build a searchable online reopsitory of PUBLICLY AVAILABLE information (forms, reports, photographs) related to arrest (and other publicly available law-enforcement-related) records.
This is inspired by Matt Van Swol's (@mattvanswol on X.com) work on investigating such records in North Carolina. Large scale misclassification of arrestees' race seems to be occurring. The scale is such that the misclassification seems interntional. Potential purposes are propaganda, manipulation of government funding mechanisms, and Cloward-Piven overload.
I'd like Roberto to
- inventory governmental and academic sources of primary data
- determine how to obtain information, and how to minimize costs of doing so
- determine a way to vet the authenticity of the data
- measure the accuracy of the data
- analyze relation of inaccuracies to particular governments, offices, and personnel
Once retrieved as public information, I suppose it could be hosted and exposed via API or app-y interface?
I suppose there might well be other fonts of governmental or institutional records that might be similarly addressed.
I think these efforts would increase transparency, combat subterfuge and fraud, and increase accountability.
Any thoughts, all?
@BrianRoemmele
1
32
As cleaned up by Gegorio
-
I want to task one of my Grok Bots (Roberto the Record Ranger, reporting to Gregorio) with building a searchable online repository of PUBLICLY AVAILABLE law-enforcement records: arrest forms, reports, and related documents.
This was inspired by @mattvanswol's work on North Carolina records, which suggests large-scale misclassification of arrestees' race. If the errors are as widespread as they look, it's worth asking whether they're systematic, and the data itself should be able to show that.
I'd like Roberto to:
inventory government and academic sources of primary data
find out how to obtain the records lawfully and as cheaply as possible
vet the data's authenticity
measure its accuracy
analyze whether the inaccuracies cluster by jurisdiction, agency, or office
The results could be hosted and opened up through an API or a simple app. Privacy comes first: aggregate and anonymize where possible, leave out expunged and juvenile records, and do no mugshot shaming.
The same approach could work for other government and institutional records too. The goal is more transparency, less fraud, and real accountability.
Thoughts, all? @BrianRoemmele
20
Musk and the others have 'gotten richer' by making the companies they run worth more.
They are not diving into piles of gold coins, or sewing Benjamins into mattresses or baking them into pies. The companies they run, and that others also own much of, are MORE VALUABLE.
Your envious grumpy shouting is SOOOOOOO 1980's 'Spartacus Youth League'
Go stand in the corner, facing away, while productive people make life better..
1
32
Once again, birds aren't real..
B iological
I ndependent
R emote
D etection
S ystem
WAKE UP YOU GUYS
15
The 'AID' in 'USAID' stood for 'Agency for International Development', surreptitiously taking advantage of the pronunciation to gull well-meaning Americans.
Dead Kennedys had a line
'The Peace Corps builds up the labor camps /
While they think they're building schools!"
streetdirectory.com/lyricadv…
Brian,
You are likely familiar with the late Wayne Greene. I believe he started Byte Magazine. I once bought a pamphlet on a 'Plant Growth Stimulator' functionally equivalent to a blood purifier gizmo based, I think, on the patent at the Albert Einstein.
Related: the 'Beck Protocol'
bobbeck.com/pdfs/build-your-…
26
When I think about it, I bet Venn is kind of angry that the one thing he is best known for is a pair of circles.
2
85
Nerds, civilized tech-heads, and lessers!:
Grab a cool handle!!
local.ai/georgetaylor/invite
86
Just gotta say:
'numpy' rhymes with 'frumpy'
'scipy' rhymes with 'hippie'
#python
smdh
65
#NerdAlert
I follow @BrianRoemmele here on X, and subscribe to his website readmultiplex.com. Those interested in insights into fascinating technology, artificial intelligence, robotics, foreseen changes in the economy, ane so much more, I suggest following him here, and visiting, and hopefully subscribing to, his website.
Brian has created a Zero-Human Company, an organization for which he serves as chairman, Grok serves as CEO, and a number of other AI agents serving as COO, CTO, CMO, and MANY (hunbdreds) of other roles.
Trying to follow, I have been able to get an agent (named 'Withnail') comfortably housed,
Raspberry Pi 5 (8GB)
Hermes 0.19
Ollama, using qwen2.5-3b-64k
Telegram
This DOES WORK. So, I am happy and want to expound.
I've created, and have been using, a knowledge base in a repo, a copy of which is on GitHub.
I've been working with Grok to smoothify information acquisition, summarification, and use, based on a Wiki like that discussed by Andrej Karpathy, with an ontology layer above the info layer.
If any of this makes sense:
@brianroemmele - Brin Roemmele
@karpathy - Andrej Karpathy
@garrytan - Garry Tan
@bOpen_AI - started by Kurt Wuckert and his team at GorillaPool
77





