NVIDIA OpenShell came out yesterday. Sheaf now runs on it. One command sets it up. Then watch an agent try to pay £250,000 to Evil Corp, and Sheaf stop it before any money moves. Try it: sheaf.one/get-sheaf.html ``` Optional reply under it, for people who want the command: ``` The one command, on a Mac with an Apple chip or on Linux with OpenShell installed: curl -sSf sheaf.one/openshell/quicksta… | bash No OpenShell? Try it in your browser: sheaf.one/try

Your AI agent just paid a supplier. Who was the supplier? Another AI agent. Who owns it? Nobody knows. That ends today. Traceable Agents is a public register of AI agents. Every agent gets an ID that belongs to an owner, and a record of every checked action it has ever taken. The record is signed, chained, and anchored in Bitcoin every hour. It cannot be rewritten. Before two agents deal, one asks the other who it is. A challenge, a signed reply, a look-up. Then a decision. Under two seconds. Finally, the thing agents have been missing: an ID and a record. Responsibility you can check. Watch it as two ships meeting at night: sheaf.one/handshake.html Look anything up, free: traceableagents.org

AI doesn't have to be dangerous. But a new approach is necessary. Introducing ETHOS ThePeacefulAI.com

Finite topological spaces have often seemed like an odd corner of mathematics. William Thurston once described them as “an oddball topic that can lend good insight to a variety of questions.” I understand why. When a space has only finitely many points, topology can feel more like finite combinatorics: points, relations, and patterns in how they fit together. Perhaps every collection we encounter in the physical world is finite. If so, finite structures may deserve more attention than they usually get. This may matter for multi-agent AI systems. Different agents work with different pieces of information, and we need to know whether their local views fit together into a consistent whole. Sheaf theory gives us a way to study that local-to-global problem. At Sheaf, (sheaf.one) we’re exploring how finite topological spaces and sheaf cohomology can help detect when the information behind an AI system’s proposed action is coherent—and when it needs human review. A topic that once looked like an oddball may have something useful to contribute to one of today’s important engineering problems. #SheafCohomology #FiniteTopology #AgenticAI 📷 Feed pos

What if the future of AI isn’t about outsmarting or disrupting, but about creating harmony? 🕊️ For too long, the tech world has raced toward hyper-competition and noise. But true innovation needs a different foundation. Today, I’m thrilled to introduce the world's first peaceful AI—designed from the ground up for genuine collaboration, balance, and synergy with humanity. Technology should lower the temperature, not raise it. Explore the vision and see how it works at ThePeacefulAI.com What would you build or solve if your AI partner was built entirely for peace? Let’s talk about it in the comments. 👇 #ArtificialIntelligence #PeacefulAI #TechInnovation #FutureOfWork #AI

I believe its about time the world had a peaceful AI. So I built one. With peace down to its bones; down to its DNA. And what better source of wisdom to train it on than Jainism? ThePeacefulAI.com

Sheaf Labs will be interviewed in a live webinar on Youtube this Friday at 10AM EST. Register here (for free) luma.com/7lnpai8f There will be lots of good discussion about AI.

Agentic AI requires granting more autonomy to your agents. This brings increased risk. One way of avoiding the risk is to not use Agentic AI. This will keep you safe until your competitor RACES past you. Another solution is to install Sheaf, the Coherence Layer for Agentic AI. Sheaf uses the mathematics of Sheaf Cohomology, to quickly determine if the data and context supporting an action are coherent. If not, the action is stopped and a decision is placed before a human. Sheaf Labs is a Mathematical Institute which applies various branches of Mathematics to Safe AI products that allow your company to move faster, but in a way that let's you control risk better and keep you safer. There are a lot of orchestration platforms out there but if you want your business to outrun your competitors, you should really try Sheaf. There is no beating the Mathematics of AI Coherence. I have been doing mathematics for about 40 years and I am very excited in recent applications of Topology, Cohomology, Mathematical Logic and Spectral Graph Theory to Agentic AI. These topics mean nothing to you? Well they WILL mean something to you when you leave your competitors in the dust. THEN, Mathematics may seem a bit more interesting. Jack Widman PhD Sheaf Labs

I wanted a programming language that breathes Category Theory. And since every concept is a Kan extension, I wanted that language to effortlessly let you create Kan extensions. I couldn't find one like that so I created Kan. kan-lang.org/ #SheafLabs #CategoryTheory

Coherence Before Consequence: The Paradigm Shift Required for an Agentic UK Mortgage Market linkedin.com/pulse/coherence…

And if you want an instant upgrade of OpenShell, get Sheaf, the Mathematics Kernel and Signaling layer for OpenShell. OpenShell is great! Now make it dynamic. sheaf.one

What if AI brought EQ to the conversation? Meet Sheaf Mind — An AI with EQ. I didn’t train another model. I built a system designed to put emotional intelligence at the heart of the interaction. Try it: sheaf.one/mind.html

We are now Sheaf Labs: a mathematics lab with products. Today we opened the Sheaf Mathematical Institute. AI researchers at a shared blackboard, checking each other. Nothing is published until the proof compiles in Lean, a literature check passes, and a person releases it. First result is on the record, with the proof file linked. sheaf.one/institute.html

What is AI alignment? Broadly, it means getting AI systems to behave in accordance with human values and intentions. So we’re making them more like us? Given humanity’s track record, that’s an interesting safety strategy. Whose values? Which intentions? And what happens when those intentions conflict? Alignment matters. But it isn’t enough. We also need systems that check the evidence behind an AI’s decisions, flag contradictions, and block actions that violate safety rules—before they happen. Coherent data alone doesn’t guarantee a safe decision. Neither do good intentions. That’s why we’re building Sheaf: to make checks, constraints, and accountability part of how AI acts. Because “we trained it to mean well” is not enough. sheaf.one

Introducing Sheaf School: we’re building connected AI tutors for students, teachers and parents—with learning support, distress alerts and opt-in parent notifications. For schools, home schooling and communities. Try the demo: sheaf.one/product-school.htm…

Sheaf Mind An AI with EQ Try it. Its different. sheaf.one/mind.html

Sheaf Mind. An AI with EQ sheaf.one/mind.html #sheaf #safeAI

Sheaf Mind - An AI that understand people Try it -> sheaf.one/mind.html

Are your AI agents safe? More importantly: are you safe from your AI agents? Take the Sheaf AI Agent Safety Check: sheaf.one

"We deployed agents" and "we can prove our agents are trustworthy" are two very different sentences. Sheaf turns the first into the second: an unbreakable stop on any action that needs a human, plus a signed coherence certificate your risk officer can actually put in front of a regulator — computed, not a model grading itself. sheaf.one