Security, AI, and strategy in motion.
Unsupervised Learning follows the places where AI, security, infrastructure, and human judgment collide. Its standard episodes move through current signals and personal working notes, while conversation episodes linger on system design, security leadership, and applied AI.
PRESS PLAY
Unsupervised Learning is a cybersecurity and technology analysis show with AI as its recurring pressure point. It is not only about tools. The show keeps asking what changes when AI becomes part of hacking, software work, research, prediction, security operations, and personal productivity. The standard episodes have a rapid, notebook-like rhythm. They move across security headlines, platform news, infrastructure risk, and the speaker’s own experiments with coding agents, newsletters, essays, videos, and workflows. That loose surface still has a clear center. The show is interested in leverage: where a new model, chip plant, exploit pattern, or automation loop changes what people can do. Recent material points repeatedly to AI malware, AI-assisted hacking, SaaS as an attack vector, a Chrome 0-day, critical infrastructure attacks, and the question of whether AI skepticism is missing something important. National-security concerns sit close to the technical ones. Taiwan, TSMC, Arizona semiconductor manufacturing, undersea cable drones, Ukraine’s asymmetrical attack, and Iranian critical infrastructure attacks all appear inside the same field of view. The show also makes room for arguments about meaning and self-management. Stoicism, relationships, sleep experiments, productivity bursts, prediction markets, and the phrase “post-AI world” are not side decorations; they are part of how the show frames readiness. Conversation episodes give the feed another shape. Michael Brown discusses AI and machine-learning security research, Buttercup, and the AI Cyber Challenge. Christine Gadsby from BlackBerry talks through AI in cybersecurity, security leadership pressure, outsourcing, boards, incident response, and threat hunting. Marcus Hutchins appears as a reference point in an AI debate, especially around skepticism, hype, scams, and malware research. The tone can be blunt and informal, but the show’s core appeal is analytical rather than theatrical. It suits a listener who wants technical signal mixed with strategic speculation and personal operating principles. It is especially useful for someone tracking AI’s effect on security work, software practice, and the wider systems that technology depends on.
Made for: Suits a listener who follows cybersecurity and AI closely and wants technical signals connected to infrastructure, national security, and practical adaptation. Best for someone comfortable with fast shifts between news, tools, arguments, and personal operating models.
What sets it apart: The show links hands-on AI and security concerns with wider strategic systems, from semiconductor manufacturing to prediction markets and critical infrastructure. Its edge comes from treating readiness as both a technical and human problem.
Unsupervised Learning is about ideas and trends in Cybersecurity, National Security, AI, Technology, and Culture—and how best to upgrade ourselves to be ready for what's coming.
A suspicious political image gets its shadows checked, then your trust in visual evidence gets checked too.
Press play if you want
Short reviews from the PlayNext desk, based on the episodes we processed.
A thoughtful, wide-ranging episode on TSMC, Taiwan risk, and the habits that help make uncertainty feel manageable.
Episode 465 brings big technology questions back to human scale, moving from a rebuilt personal website and a podcast recommendation to the host's decision to buy more TSMC stock.
Rob Allen turns zero trust into a practical discussion of application, data, and network controls.
Rob Allen makes zero trust feel less like a slogan and more like an operating model. He walks through ThreatLocker's denied-by-default approach to endpoint protection, with attention to which applications can run, what those programs are permitted to do, how sensitive files are…
A grounded way to tell when AI labor savings vanish into human cleanup.
Unsupervised Learning treats AI job replacement as a workplace problem, not a race over model size or technical bragging rights. Daniel Miesler lays out levels of AI workers based on how they hold up inside everyday messy systems: awkward interfaces, prompt-style instructions,…
A clear tour of AI-enabled scams, election influence, intelligence gathering, and privacy risks around everyday technology.
Unsupervised Learning frames early AI breaches as a broader security story, moving from pig-butchering scams and exploited lawful-request backdoors to foreign operations using AI to shape political opinion and interfere in elections.
Convincing AI images turn into a question of evidence once crime, politics, and escalation enter the frame.
AI's most unsettling frontier is framed as a problem of evidence. The host starts with synthetic images persuasive enough to fool almost anyone, then follows the concern into criminal investigations, politics, and the possibility that a fabricated threat could trigger real-world…
A security-minded tour through CrowdStrike, tech policy, merit, and the arguments underneath current events.
Unsupervised Learning uses the CrowdStrike incident to open a wider technology conversation, with Lee Honeywell's point about a bad software update being hard to distinguish from a cyber attack giving the episode its cleanest frame.
Technology's mechanisms, power's consequences.
Dave, Joe, and Maria break down phishing kits, deepfakes, and social-engineering scams weekly.
The host discusses tech trends, including Google I/O and Apple WWDC, and shares personal experiments with noise-canceling earbuds for sleep.
Daniel Miesler debates Marcus Hutchins about the value and hype of AI in cybersecurity.
Daniel Miesler describes using Claude Code across five repositories to repair thousands of blog posts, migrate images to Cloudflare, and automate work he estimates would have cost thousands of dollars.
Daniel Misour reviews toolchain changes involving WhisperFlow, Zyte, and the Firefox-based Zen browser, including privacy tradeoffs and workflow consolidation.
Daniel Miesler analyzes the CrowdStrike outage, arguing that a bad software update can look indistinguishable from a cyber attack and criticizing the company's initial public-relations response.
Rob Allen, Chief Product Officer at ThreatLocker, explains why the company uses zero trust to block everything unless it is explicitly allowed.
Christine Gadsby of BlackBerry argues that AI is already useful for incident response, threat hunting, indicators of compromise, and automated security decisions.
Unsupervised Learning’s “The First AI Breaches” centers on the moe?AI database breach, where chat history, user accounts, and personal addresses were exposed.
Daniel Miesler argues that AI model winners will be decided by four pieces — the model, post-training, internal tooling, and agents — not by parameter count alone.
Daniel’s RAID framework, short for “Real World AI Definitions,” turns terms like AI, RAG, agent, AGI, and ASI into plain-language thresholds tied to work, data, and capability.
An interview with Michael Brown about using AI and machine learning to solve cybersecurity problems and his team's success in the AI Cyber Challenge.
The host discusses changes to the podcast format to speed up content delivery and introduces a new video on the future of hacking.
Daniel opens UL No.