Monero proved privacy works. @Zano_project is about where privacy goes next. Here’s how the two ecosystems differ and why that matters. 🧵

With ATLAS, one identity can move across social, business, commerce and financial services—creating a connected journey: Discover → Buy → Pay → Settle → Operate One identity. Connected commerce. One ecosystem. #NexarNetwork #ATLAS #Web3

gm, Saturday edition. No roadmap talk today. Just this: what got you into decentralized AI in the first place?

Something bigger is taking shape. ⚡ Nexar × ATLAS One ecosystem. One connected experience. Don’t just watch it. Enter it. 🟡 Explore ATLAS 👇 nexarnetwork.org/atlas #NexarNetwork #ATLAS #Web3 #BuildTheFuture

What if your content could do more? 👀⚡ Create. Go LIVE. Grow your audience. Receive gifts. Earn NXR. Build your personal brand. This is the Creator Economy inside ATLAS. 🟡 Your audience. Your content. Your opportunity. 🔗 Enter ATLAS⁠� #ATLAS #NexarNetwork #CreatorEconomy #NXR #Web3‌‌

Multi agent AI challenge is not getting the agents to talk to each other, because they already can. Knowing the decisions they have made, the information passed around and things changed in the process is where the challenge lies. It becomes important very quickly in a business, you don't just need the final answer. You need to know why something happened and the correctness of the information. We'll need much better ways of tracing AI decisions as these systems become more autonomous.

Robots are going to expose a lot of assumptions made about AI. I'm interested in what happens when foundation models have to operate continuously in the physical world.

The answer to AI safety is not simply more rules, the answer is rules are useful. If a company can say its model passed an internal safety evaluation and that's enough to deploy it, we haven't really solved the issue. But if the alternative is a regulatory system so expensive that only the largest labs can participate, we've created another problem. It would be great to see independent testing become normal. Tell us what was tested. Tell us what failed, what was fixed, and let someone outside the company verify it. That gives people something better than either blind trust or endless arguments about whether AI is dangerous.

I was reading the report of @amazon exploring a deal to move roughly $8 billion worth of Nvidia chips into a special purpose vehicle and then lease the infrastructure back. AI compute is infrastructure that can be financed, leased, and owned by different parties. Amazon could potentially have investors own the chips while Amazon continues operating them in its data centers. They get access to the compute it needs without having to keep the entire cost of those assets on its balance sheet.

gm and happy Friday. Somewhere on the mesh, a node is running inference that never touched a hyperscaler. That is the win.

Private AI for patients. Records never leave the building. Local AI on the PowerNode keeps care where it belongs. #AI #DePIN #PowerNode #PrivateAI #LocalAI

The Founder is now inside ATLAS. 🟡 This isn’t just a profile. It’s the beginning of a new way to connect People, Business, Commerce, Payments & Technology in one ecosystem. Meet Mahmoud Elgabry — Founder & CEO of Nexar Network, directly from inside ATLAS. . nexarnetwork.org/atlas/netwo… . One ecosystem. One identity. One connected experience. Welcome to ATLAS. ⚡

The PowerNode is here. Now shipping. Order yours at pai3.ai. #AI #DePIN #PowerNode #PrivateAI #LocalAI

lights out. inmate requested his letter board back to update it to 'break all time highs.' request denied. JAIL JACK $JACK is going for the ATH instead. different board, same ticker.

exhibit A, sept 27: "Never selling my Jack Doherty coin holding forever." exhibit B, oct 1: "I'm never selling… you guys have my word." the defense rests. JAIL JACK $JACK 🔒

We are going to have a lot more software than we know what to do with. In many ways, it is a good. Things that used to require a large engineering team will be possible for a much smaller group, and some ideas that never made economic sense to build will become viable. But abundance changes the nature of the problem. When software becomes cheap to create, judgment becomes scarce. It's important because software doesn't stay on a screen. It ends up in hospitals, banks, factories, schools and public infrastructure. A bad product can be annoying. A bad system can affect millions of people. Software engineering is going to give us extraordinary leverage. What we do with that leverage is of great importance.

Where does your data go tonight? Nowhere. It stays home, locked on the PowerNode. #AI #DePIN #PowerNode #PrivateAI #LocalAI

A lot of AI development has happened around a relatively small number of languages and datasets. So when you take the same model into another part of the world, it may understand the words but miss the context around them. That is why I find regional models interesting. Not because every country needs to build its own frontier model, but because there are things that simply aren't going to be captured properly by a system trained somewhere else. Language carries history, culture, local expressions and ways of describing things that don't translate neatly. If AI is going to become part of everyday life everywhere, representation in the training data is going to matter just as much as model size.

The worrying part of an AI agent trying to hack a government website is not whether it succeeded. It is that the attempt can happen at all. A human researcher usually has to decide what system to target, what information to look for, and what action to take next. An autonomous agent can turn those decisions into a sequence of actions and keep going with very little supervision.

gm. New month, same thesis: whoever owns inference sets the rules. Let's work.