Excited to talk to three heroes of mine, all in one sitting! We will explore how tools, both surveillance and privacy tools, can evolve beyond their intended use cases - as they have before. Imagine a different future: what if we embraced surveillance, in exchange for security? What would that world look like? Join us on the first livestream on @ethereum @VitalikButerin @Ada_Palmer @SherriDavidoff

I found this article by @ccatalini in @HarvardBiz fascinating, as it touches on a sensitive topic today - companies giving away their proprietary know-how to AI companies, creating powerful new competitors. But as someone who doesn't spend all day researching AI, I found it hard to follow - as would CEOs of non-tech companies. I made AI make some infographics to translate it to actionable C-suite info. TL;DR: don't give away your business secret sauce to AI companies like OpenAI, Anthropic, Google. Use it to create more value for yourself! These infographics show how to do that: 1. The team you need 2. How evals work 3. Example: evals for life insurers

The Hazards of Over-Automating according to @ccatalini 1. Flattening Divergent Views ("Decision Slop"): Tools that automatically summarize context risk compressing nuanced expert hesitations into premature consensus. 2. Eroding Internal Expertise: If experts are cut off from real-world friction and operational decision-making, their judgment degrades. During sudden market shifts, models will be confidently wrong and no internal experts will be left to catch the flaws. 3. Ceding Competitive Advantage: Closed AI labs aggressively acquire expert traces and specialized data. Firms that hand over their operational logs to closed model providers risk becoming thin wrappers around intelligence they no longer own.

X is an AI bubble. Spend enough time here and you’d think every company is ready for agents managing other agents. Meanwhile, plenty of businesses are still trying to get one workflow running reliably without someone manually carrying it through every step. The gap goes well beyond teaching an agent how the company works. Someone has to decide which problem is worth solving, understand why the current process exists and work out what should change when AI takes on part of it. That requires product judgment. An agent can produce a good answer while leaving the team to export the data, move it between systems, chase approvals, correct errors and finish the job. Those steps belong in the scope. Otherwise, you can make the model’s part faster while barely changing the work around it. This is why I want engineers working directly with the people who run the operation. Watch how they handle real cases. Ask why they override the documented process. Put a working version in front of them early enough to discover that you’ve misunderstood something. Then do the engineering required to make it dependable. Connect it to the right systems. Test against cases the business has checked. Enforce what it can change. Make failures visible and recoverable. Keep checking it as the software and business rules change. Context and skill files help capture the knowledge. Evals check whether the system applies it correctly. The integrations and controls let it do useful work. And someone on the business side needs the authority to change the process, get the team using it and own the result. That responsibility cannot sit permanently with the engineer who built the first version. I’d measure progress by how much work gets completed correctly, how much human effort remains and whether the client can keep improving the workflow after we leave. Multi-agent orchestration can be part of the solution. But the company still has to do this work. The client should finish the engagement with a better operation, evidence that it works and people who know how to run it. That's what we provide at @LimestoneHQ.

“Tools that summarize documents, interactions, and context risk flattening divergent views into premature consensus” >> I feel this a lot

Incredible video shows a Russian drone attempting to destroy a Ukrainian ground robot but being disabled when the robot deploys an anti drone net and catches it. We literally have robots fighting each other on Ukraine’s frontline.

ELI5 of a key security mechanism in Signal

Doing a bit of a self-experiment. Goal: use my personal health and travel data to provide personalized diet and exercise recommendations for me, using frontier models but in a way that avoids leaking to them any private information. Strategy: use a local model (Qwen 3.8 Flash Next) to orchestrate, and use powerful remote models as a tool call to benefit from higher-level thinking and knowledge that the local model does not have. Three-layer approach to privacy: Avoid leaking personally identifiable information, or leaking my identity through writing style -> local model writes the queries to the frontier model, not me Avoid leaking who I am through the payment channel -> zkAPI Avoid leaking who I am through networking / IP -> Tor You need all three (and finally we have all three, at least to some extent) A skill file teaches the local model when and how to construct minimally-data-revealing requests to remote models. Use zkAPI-wrapped-with-Tor as a CLI tool. And everything works! I got the recommendations back, info from frontier models helped to improve them. Main deficiencies: * Tor is really not optimized for request-by-request de-linking, which is the only form of network-layer privacy that really makes sense in today's world (long-running identifiers are too fragile). Probably not private enough, and latency 10-100x higher than it could be, at the same time. * The skill file's request construction strategies are definitely far from optimal. * Qwen 3.8 Flash Next is still too slow for comfort. It's comfortably running at 20-30 TPS, but it would only really start to feel fast at 100+ * There is a tradeoff: the more careful you are about what data you give to a remote model, the less it can help you github.com/ethereum/zkapi/pu…

Last Sunday, Vitalik released the full version of Snowmoon. An underpinning concept to many of the ideas in the story is something V calls “Coordination Quotient.” Coordination Quotient is his invention, there’s no pre-existing literature on that exact idea! In simple terms, CQ is a cure for the prisoner’s dilemma or tragedy of the commons. A community with high CQ is capable of defying its respective social or economic nash equilibrium. High CQ basically says that even if we are individually incentivized to act selfishly, we are effective at *choosing* to act in the collective interest without external forces pressuring us to. My personal analogy is that a society with a perfect Coordination Quotient wouldn’t need formal laws or a governing body to function at scale because it would be so good at acting in good faith and achieving sophisticated positive-sum goals.

Paradigm GP @danrobinson says AI labs are trapped in a prisoner’s dilemma where everyone would rather slow down, but nobody wants to fall behind first: "There is this kind of prisoner's dilemma-like component to AI pacing, where you might have every participant in an AI race prefer to, at least on the margin, go slower than they are going." "A lot of people at these labs would prefer to slow down, but are concerned that if they do, others won't." "You would rather that both you and your competitor go slower. But when you're playing, you say, 'Okay, if they're going to go ahead, then things are best if I'm still following them.'" "Both speeding ahead and pacing the frontier are equilibria. A lot depends really on these contingent factors about what happens." @paradigm

I think being able to video chat with a cute woman who never gets tired of you is going to be a huge disaster for our already loser-based society

We are now entering into an era where any product can be created exactly to a consumer's preferences & needs. I vibe-fabricated a dog door with a wifi-controlled lock, perfectly to the specifications & design of my house. I know nothing about metal fabrication or electrical engineering. It's now getting manufactured and delivered in 2 weeks -- for almost the same cost if I bought a mass-produced item off-the-shelf.

As a harness builder I am both amazed and terrified by agentic computer use. I get why it's compelling, but the fact that the industry seems to be yolo'ing into it gives me great pause. We're telling the models to use products that were built for humans. Those humans have survival and social instincts that prevent them from causing harm. Don't go on a spending spree on amazon. Don't reveal this photo that would embarass your friends. As amazing as modern models are, I don't think they understand these human instincts well yet. With these tools, you're putting your identity and reputation on autopilot. It's very hard for me to see how this current version of it doesn't end badly. I have a few computer use setups of my own (carefully firewalled), and when I see these agents "use" a computer, I'm quickly reminded that we are dealing with alien beings. Don't assume they think like you and always have your best interests at heart! They don't even have a heart to begin with!

🤣

Sanctuary Technologies: resilient tech on display at the Singapore @ArtSciMuseum @GrapheneOS on display! Along with @azonenberg’s “fab in a shed”

I'm not Catholic, but i agree with the pope. More than trying to say "human art and machine art are different," I think he is actually trying to say "humans and machines are different."

NEW: According to a bombshell report in the New York Times, Anthropic co-founder Chris Olah threatened to walk out of Pope Leo XIV’s AI encyclical launch in May because the pope rejected the idea that machines can be conscious. Olah’s team then privately lobbied the pope’s advisers “to take the possibility of model consciousness seriously.” Pope Leo XIV held firm. For months, Anthropic has wined and dined theologians and religious scholars under nondisclosure agreements, hoping they would bless the idea that Claude has moral standing. thelettersfromleo.com/p/nyt-…

September belonged to crypto x artificial intelligence (AI): $NEAR up 183% $VVV up 70% $WLD up 47% $TAO up 37% Grayscale's Artificial Intelligence Crypto Sector returned 54% for the month, versus 24% for the broader crypto market. The move may reflect growing interest in blockchain as potential infrastructure for AI. Read more on The Stack here: grayscale.com/the-stack/the-… Learn more about the Artifical Intelligence Crypto Sector here: institute.grayscale.com/cour…

Join us in Singapore on Monday to talk about Zcash going institutional!