🧠 Unsupervised Learning NO. 542 is out. 🔓 AI attacker teams, 🤖 model routing, 🧭 TELOS risk registers. 🙏 Thanks to our sponsor Sublime: https://ul.live/sublime_n542 https://newsletter.danielmiessler.com/p/unsupervised-learning-no-542
ᴅᴀɴɪᴇʟ ᴍɪᴇꜱꜱʟᴇʀ
@danielmiessler.bsky.social
AI / Security Researcher and Entrepreneur. Founder/CEO of Unsupervised Learning. Building AI that upgrades humans.
I built an AI agent called Socrates that refuses to give answers. It asks you for yours, then asks how you know. For kids doing homework, and for anyone who can feel their thinking getting lazy. One file, free: https://danielmiessler.com/blog/the-socrates-agent
What I think is really important right now is making sure you're not living someone else's life.
New post: Why I Hate Anthropic and You Should Too. Because clearly the company spending billions on safety, pushing back on the Pentagon, and warning us about AI is the real villain here. https://danielmiessler.com/p/why-you-should-hate-anthropic
One issue I'm having: knowing when to pivot between improving the system vs. using it to do work. It's never been more enticing to sharpen your tools. But you can lose days on that instead of cutting the log. Message for myself. Maybe you too.
My 2026 cybersecurity predictions: The primary question becomes how good your AI is vs. attackers' AI. Agents will finally make asset management possible. Security certs plummet in value. Junior talent faces massive pressure. https://danielmiessler.com/blog/cybersecurity-changes-2026
Just read. One book at a time, or 17. I’m currently reading: Letters to a Young Poet, The Dharma Bums, Accelerando, The Peregrine, Dominion, and more. It’s chaos. I’m embracing it.
One good way to think about AI—and all major tech movements like electrification, the internet, etc—is as a Technological Weather Event. Not just to say that it’s mostly unavoidable, but that it really matters how we think about and prepare for them as a population.
You don’t have to believe that companies want to fire all their employees to see AI’s threat to jobs. That’s negative framing. Most company leaders are regular, decent people who generally care about others. danielmiessler.com/blog/better-...
A Better Way to Think About AI Job Replacement | Daniel Miessler
SECURITY | AI | PURPOSE :: Building AI that upgrades humans.
danielmiessler.com
The future of hacking isn’t smarter tools—it’s better context. Why defenders (and attackers) will win by building real-time world models. 📽️ New video: youtu.be/UwTTcka1Wd8
The Future of Hacking is Context
YouTube video by Unsupervised Learning
youtu.be
We're about to get precisely the right amount of AI Security. No more and no less. Hear me out… I talked before about what I called the Efficient Security Principle, which says the excitement for a technology lowers its required security bar... Read more here: danielmiessler.com/blog/ai-secu...
We Can't Really Affect AI Security | Daniel Miessler
SECURITY | AI | PURPOSE :: Building AI that upgrades humans.
danielmiessler.com
With all this AI job and economy disruption, don’t forget that we tend to overestimate short term progress and underestimate it in the long term. Read the full essay to see what I think it will look like: danielmiessler.com/blog/slow-th...
AI Job Replacement: Slow Then Fast, May 2025 Edition | Daniel Miessler
SECURITY | AI | PURPOSE :: Building AI that upgrades humans.
danielmiessler.com
🚀This week's UL is live! Featuring: 🧠 AI finds a 0-Day 📦 Postman leaks secrets 📹 Unified Entity Context explainer 🔓 GitHub MCP leaks ⚔️ Google vs. OpenAI vs. Apple (AI vision) 💡 High Agency as a mental model …and more newsletter.danielmiessler.com/p/ul-482
Unsupervised Learning NO. 482
AI Finds an 0-Day!, Postman Leaking Secrets, High Agency Mental Model, My Unified Entity Context Video, Github MCP Leaks Private Repos, Google vs. OpenAI vs. Apple on AI Vision, and more...
newsletter.danielmiessler.com
Watched The Accountant 2 and loved the concept of Acquired Savant Syndrome—when brain injury unlocks genius. What if ASI could figure out how to trigger that intentionally? AI shouldn’t just make better AI. It should make better humans.
The future isn’t AI plus context—it’s context first, then AI. I call it Unified Entity Context, and I think it changes how we approach everything from security to decision-making. New video: youtu.be/IHUqk90ch7I Sponsored by Project Discovery → ul.live/PD1
Unified Entity Context
YouTube video by Unsupervised Learning
youtu.be
🚀This week's UL is live! Featuring: 🎧 AirPod cameras? ⚙️ Systems > tools 📚 Can AI boost education outcomes? 🔐 VC data breach 🌐 Claude gets web search 🎤 Speaking at Nahamcon + RSA recap w/ @jhaddix ...and more: newsletter.danielmiessler.com/p/ul-481
Unsupervised Learning NO. 481
AirPod Cameras?, Systems over Tools, AI Improves Education Outcomes?, VC Data Theft, Claude API Gets Web, Speaking at Nahamcon, and more...
newsletter.danielmiessler.com
Most every task is easier if you have more information. One of AI's primary benefits won't be just making smarter decisions itself, but providing the narrative context to make the decision easy for anyone.
Our jobs mentality is based on a lie. ➡️ We were told that there are special people who make things—but we're not one of them. ➡️ And our role is to work for one of them, hopefully… Fuck that. The illusion is collapsing. You are the special person. You can make things.
What if AI enables megacorps to become giant incubators of millions of startups? - Identify problems anywhere - Find the best solution - Make a better solution - Distribute using your network How do regular startups beat that? danielmiessler.com/blog/ai-solu...
AI Solution Factories | Daniel Miessler
SECURITY | AI | PURPOSE :: Building AI that upgrades humans.
danielmiessler.com
🎙️ I sat down with Bar-el Tayouri of Mend AI to talk about the future of AI security—and why it’s evolving faster than any category before it. We covered malicious models, agent discovery, and more. If you're building with AI, this one's essential. youtu.be/538JTah7dTY
A Conversation with Bar-El Tayouri from Mend.io
YouTube video by Unsupervised Learning
youtu.be
I’ve been thinking about how AI might reshape what it means to be a developer. We’re hearing a lot of predictions—some saying tech jobs will vanish, others saying AI will supercharge demand. But I think the more interesting shift is in who companies will need. 🧵
One of the biggest use cases for prompt injection and persuasion will be in product API’s targeting models and digital assistant agents with marketing. Marketing targeting humans won’t be nearly as important as the marketing targeting the AI working for the humans.
🚨 This week’s UL is live! Featuring: 🚢 Supply chain disruption incoming 🗳️ Easterly speaks out on cyber politicization 🖼️ Images now in the OpenAI API 🧠 Context + AI = Enterprise AI 🧾 Microsoft Recall returns …and more 📬 newsletter.danielmiessler.com/p/ul-479
Unsupervised Learning NO. 479
Supply Chain Crash, Easterly Asks for Opposition, Images in the API, Context+AI=EnterpriseAI, Microsoft Recall Returns, and more...
newsletter.danielmiessler.com
Here's something nobody's talking about with AI... Removing the Fog of War. Elite generals and principal engineers shine when info is scarce. Expertise = stitching limited info into a usable world model. 🧵
🚀 This week’s UL is live! Featuring: 🔓 Pliny’s jailbreak repo 📜 tRIOT—now in blog form 🏥 Health data leaking via Google Ads 👁️ Protester surveillance tactics 📺 How YouTube won the UGC war newsletter.danielmiessler.com/p/ul-478
Unsupervised Learning NO. 478
Pliny's Jailbreak Repo, tRIOT as a blog post, Health Data Google Ad Leak, Protester Surveillance, How YouTube Won UGC, and more...
newsletter.danielmiessler.com
Moats aren’t just for businesses. In the age of AI and infinite replication, individuals need them too. Ask yourself: — What makes you hard to copy? — What will still be uniquely you in 5 years? This is the new career planning.
💡People are obsessing over the specific AI tech that will replace human workers… …but the bigger point is that companies are so excited about replacing workers in the first place. 🔥Corporate Mindset ➡️ Independent Mindset
Visual persuasion is a double-edged sword. Once you see something, your brain starts believing it—even if it’s false. In a world of AI-generated content and visual manipulation, literacy isn’t enough. We need epistemic defense systems. How are you building yours?
💡 Want to improve your strategy? Try inversion. Instead of asking what to do, ask: 👉 What actions would guarantee failure? Then… just don’t do those things. It’s a simple but powerful shift. What are your top “don’ts” for your space?
Going to #RSA? I’ll be speaking at Aegis of Tomorrow: An AI & Security Summit on Monday, April 28 from 3–5pm. I’ll be sharing a framework for cutting through AI hype and prioritizing cybersecurity investments based on how attacker capabilities are actually evolving. 👉 Register here: lu.ma/9j1p8ixj
