Piloting The World’s First Latent Reasoning Models for Critical Systems.
Joined June 2025
- Tweets149
- Following10
- Followers5.8K
- Likes1.1K
Pinned Tweet
Here AI did not replace the mathematician.
Different tools played specific roles alongside him: exploration, proof construction, formalization, and verification, while the essential mathematical insight remained human.
Since 1995, the known frontier of “good groups” in 4-manifold topology had remained essentially where Freedman and Teichner left it. Today that frontier moves farther.
Slava Krushkal, working with Michael Freedman’s Discovery Team at Logical Intelligence, has proved that certain Sturmian groups are good, expanding the class of fundamental groups for which the machinery of topological 4D surgery theory is known to apply. logicalintelligence.com/blog…
Large language models were the first form of AI most of us met. They were never going to be the only one.
logicalintelligence.com/blog…
Logical Intelligence retweeted
LLMs are not the only way to model intelligence.
Many strong voices believe the future of intelligence lies in other routes, like energy-based modeling.
We spoke with @evelovesolive, founder and CEO of Logical Intelligence, to understand how energy-based reasoning models (EBRMs) work and what kinds of problems are well suited for them.
In this deep dive, we'll explore:
- Why LLMs struggle with problems like Sudoku
- What is an energy-based model
- How energy-based reasoning works
- What EBRMs can and cannot do
- How EBRMs handle verification and safety
Read full deep dive here:
The future is diverse, specialized intelligence.
"The future is diverse, specialized intelligence."
Today, during our inaugural AI Investors' Forum at Asia Summit 2026, our "AI and the New Tech Transformation" panelists shared insights on AI's impact on the world.
Watch all session replays: milkeninstitute.org/events/a…
#MIGlobal #AsiaSummit #Asia #AI #EconomicOutlook
Logical Intelligence retweeted
Businesses are hungry for AI but emphasizing privacy, security and transparency makes frontier models not suitable AI candidates tor mission critical industry. AI is evolving into ecosystem of smaller specified on prem models trained on customers data for businesses needs beyond LLMs
Much of the AI conversation assumes that progress means bigger models.
The takeaway across @evelovesolive 's panel at the Milken Institute Asia Summit was different: new architectures and smaller, more specialized models are the future.
Much of the AI conversation assumes that progress means bigger models.
The takeaway across @evelovesolive 's panel at the Milken Institute Asia Summit was different: new architectures and smaller, more specialized models are the future.
Logical Intelligence retweeted
Congratulations to Slava and Mike Freedman for this amazing result. Humans and AI are working together to drive the mathematical frontier foward.
As our reasoning systems become more capable, we believe that future generations of scientists and researchers will be able to tackle much deeper problems.
Over time, that same approach will unlock progress in areas like materials, semiconductors, energy, and advanced engineering, places where breakthroughs require searching huge spaces, working within hard constraints, and making sense of incredibly complex systems.
Since 1995, the known frontier of “good groups” in 4-manifold topology had remained essentially where Freedman and Teichner left it. Today that frontier moves farther.
Slava Krushkal, working with Michael Freedman’s Discovery Team at Logical Intelligence, has proved that certain Sturmian groups are good, expanding the class of fundamental groups for which the machinery of topological 4D surgery theory is known to apply. logicalintelligence.com/blog…
Since 1995, the known frontier of “good groups” in 4-manifold topology had remained essentially where Freedman and Teichner left it. Today that frontier moves farther.
Slava Krushkal, working with Michael Freedman’s Discovery Team at Logical Intelligence, has proved that certain Sturmian groups are good, expanding the class of fundamental groups for which the machinery of topological 4D surgery theory is known to apply. logicalintelligence.com/blog…
This result was achieved through human-machine collaboration.
Slava, using his “wet brain,” supplied the geometric insight and the crucial transformations that made the problem tractable. Frontier models assisted with proof construction and formalization. Our Aleph architecture then produced a Lean 4 certificate.
AI did not replace the mathematician. Different tools played specific roles alongside him: exploration, proof construction, formalization, and verification, while the essential mathematical insight remained human. logicalintelligence.com/blog…
We’re seeing strong interest from large enterprises working on some of industry’s hardest problems: material design, semiconductor manufacturing, grid management, and more.
These companies have exacting information security requirements. Adopting AI cannot mean compromising your data.
We’re building out our information security program to match and welcoming Weilin Zhong as our CISO to lead it.
The relevant question is not, “how intelligent is this model?”
We should be asking “what can this system do, what prevents it from doing something else, and what happens when those protections fail?”
That should be the starting point for any proposed deployment in a critical setting. - @PRHillmann
The “Coke vs Pepsi” race to superintelligence, consequences be damned, is frustrating to me as a builder.
My concern about LLMs is not that every improvement brings us another step toward an inevitable machine apocalypse. It is that we risk continuing to confuse increasingly impressive capabilities to mimic intelligence with real machine intelligence and giving these models increasingly consequential responsibilities.
These are general purpose models that are being smushed to fit every role for every customer. Systems are built around the model to keep them on task and staying within the constraints of a particular problem. But it will never fundamentally understand that it can’t violate a constraint. Constraint is not in its nature.
The relevant question, therefore, is not, “how intelligent is this model?” We should be asking “what can this system do, what prevents it from doing something else, and what happens when those protections fail?” That should be the starting point for any proposed deployment in a critical setting. logicalintelligence.com/blog…
Logical Intelligence retweeted
As I told the BBC, the AI labs' trust issues have been a long time coming.
They approached communications like this would be a "Coke vs. Pepsi" competition among beloved brands. Not a sea change that at best would upend jobs and at worst (apparently) doom humanity.
OpenAI boss says world 'right to be afraid' but 'should trust' AI firms bbc.in/3VgtcYO
Logical Intelligence retweeted
My interview for a senior policy role at a major lab a few years ago told me that they did not understand what was about to happen. The backlash was obvious even then. You can’t tell people that the good outcome of your thing is the loss of millions of jobs and an outsourcing of creativity, and the bad outcome is human extinction.
The frontier labs are happy to warn us that they are building something so amazing and so smart that it could threaten humanity. Let’s set aside the obvious financial motivation for saying that and take them at their word.
If they want us to take them seriously, then they need to take their responsibility seriously too. Stop building with inadequate controls, stop using the failures of those controls to promote the intelligence of your model, stop calling for a regulatory moat, and tell us what you’re actually prepared to stop doing.
logicalintelligence.com/blog…
Sofian Rhazali is joining us from Google to lead GTM.
He'll be bringing our reasoning models and agentic workflows to problems defined by constraints, especially the ones that are not easily described in natural language.
Logical Intelligence is proud to sign this letter in support of open-weight AI models.
Open-weight models expand research, innovation, and defensive security. The future is safer when defenders have the same, or better, tools than attackers.
For my first post, I’m sharing a letter @NVIDIA signed on why open models matter.
AI will transform every industry, power every company, and be built by every country.
Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.
The world needs both frontier closed models and frontier open models.
images.nvidia.com/pdf/Open-W…
See the letter with a final list of signatories here: microsoft.com/en-us/corporat…
LLMs can hallucinate. And when AI agents are writing production code, moving money, or operating critical systems, even one confident mistake can have serious consequences.
As @evelovesolive says, "If you put AI everywhere, it's gonna do crazy sh*t."