greg beier
@gregbeier.bsky.social
architecting superintelligences' judgment kernel | founder @ susarb® | gregory caldwell beier
ORCID ID: 0009-0008-3263-2447
ORCID ID: 0009-0008-3263-2447
Latest paper on SSRN
The Atlas of Accountability: A General Theory of Incentive-Compatible Semantics
Current approaches to Artificial Intelligence rely on private, opaque datasets that prioritize volume over veracity. We argue that this enclosure of training da
papers.ssrn.com
January 22, 2026 at 8:46 PM
Latest paper on SSRN
Superharddata paper now on OSF via Law Archive via
osf.io/preprints/la...
osf.io/preprints/la...
OSF
osf.io
January 20, 2026 at 11:21 PM
Superharddata paper now on OSF via Law Archive via
osf.io/preprints/la...
osf.io/preprints/la...
Now on SSRN
Superharddata: Liability-grounded information as a training substrate for AI alignment.
Superharddata: Liability-grounded information as a training substrate for AI alignment.
Superharddata: Liability-Grounded Information as Training Substrate for Aligned Artificial Intelligence
Large Language Models exhibit remarkable linguistic fluency but remain fundamentally ungrounded in consequence. Hallucination is inherent to their architecture
papers.ssrn.com
January 6, 2026 at 1:05 PM
Now on SSRN
Superharddata: Liability-grounded information as a training substrate for AI alignment.
Superharddata: Liability-grounded information as a training substrate for AI alignment.
Why do LLMs hallucinate? Training data has no cost of error. New paper proposes "Superharddata" - information produced under liability pressure - as the missing grounding layer. Physics bats last; liability ensures someone had skin in the game first.
Superharddata: Liability-Grounded Information as Training Substrate for Aligned Artificial Intelligence
Large Language Models exhibit remarkable linguistic fluency but remain fundamentally ungrounded in consequence. They hallucinate because their training data carries no cost of error. A Reddit comment ...
zenodo.org
January 2, 2026 at 5:18 PM
Why do LLMs hallucinate? Training data has no cost of error. New paper proposes "Superharddata" - information produced under liability pressure - as the missing grounding layer. Physics bats last; liability ensures someone had skin in the game first.
This is fantastic! A must watch. #susarb
Google and NASA's Quantum Artificial Intelligence Lab
YouTube video by Google
youtube.com
August 27, 2025 at 5:16 PM
This is fantastic! A must watch. #susarb
architecting superintelligences' judgment kernel
August 11, 2025 at 12:29 PM
architecting superintelligences' judgment kernel
"Many people can instruct an AI. My work is with the few who can co-create a world with one."
After months running in stealth mode, I'm ready to share what we've been building at Susarb.
After months running in stealth mode, I'm ready to share what we've been building at Susarb.
July 25, 2025 at 9:32 PM
"Many people can instruct an AI. My work is with the few who can co-create a world with one."
After months running in stealth mode, I'm ready to share what we've been building at Susarb.
After months running in stealth mode, I'm ready to share what we've been building at Susarb.
Anyone catching my vibe? #cottagetech
The Cottage Tech Revolution: Why the Future is Distributed
A manifesto for human-scale technology
gregbeier.medium.com
June 25, 2025 at 12:26 AM
Anyone catching my vibe? #cottagetech
Technology transforms when it becomes invisible. The personal computer changed the way we think. The smartphone changed how we live. Now, we are poised for another transformation—one that reclaims our most ancient capacities as we enter the era of ambient AI. #ai @appleinsider.com
The Sym-Ring — Susarb®
www.susarb.com
June 24, 2025 at 9:47 PM
Technology transforms when it becomes invisible. The personal computer changed the way we think. The smartphone changed how we live. Now, we are poised for another transformation—one that reclaims our most ancient capacities as we enter the era of ambient AI. #ai @appleinsider.com
This is the biggest technology news of the year - a very very big deal
Google Computer Solves Septillion-Year Task in Minutes
Google Computer Solves Septillion-Year Task in Minutes
www.bloomberg.com/news/videos/...
Google Computer Solves Septillion-Year Task in Minutes
Google Computer Solves Septillion-Year Task in Minutes
www.bloomberg.com/news/videos/...
Google Computer Solves Septillion-Year Task in Minutes
Alphabet’s quantum computer needs just five minutes to solve a problem that would take supercomputers around 10 septillion years. Google’s next task: coming up with an actual use for all that theoreti...
www.bloomberg.com
December 10, 2024 at 7:33 PM
This is the biggest technology news of the year - a very very big deal
Google Computer Solves Septillion-Year Task in Minutes
Google Computer Solves Septillion-Year Task in Minutes
www.bloomberg.com/news/videos/...
Google Computer Solves Septillion-Year Task in Minutes
Google Computer Solves Septillion-Year Task in Minutes
www.bloomberg.com/news/videos/...