Phoenix Cast

In this episode of Phoenix Cast, hosts John and Kyle go guest-free to unpack the latest wave of consumer AI agents: OpenAI's Dots, Meta's Muse, and Claude's newly announced Duet. Before diving in, they trade stories of AI rescuing a rotted-out grill, taming an overflowing storage room, and a three-model bake-off to design a cardboard Jawa sandcrawler for Halloween. The conversation then turns to the real price of frictionless agents: what you hand over when you connect your inbox and calendar to a consumer platform, and how Kyle's homegrown assistant "Veep" leans on enterprise agreements and local inference to stay a harder target. Along the way they get into test-driven development, agent orchestration, burning a week of tokens in two prompts, and a friendly disagreement over whether your AI should ever be allowed to delete files. If you're weighing whether to hand an agent the keys to your digital life, this one is worth a listen.Links:OpenAI Dots:https://openai.com/index/introducing-dots/Meta Muse:https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/Claude Duet (getting started):https://support.claude.com/en/articles/17631495PHP2Shell security advisory (patch your PHP):https://www.php.net/archive/2026.php#2026-10-21-5Gemini Enterprise Agent Platform (formerly Vertex AI):https://cloud.google.com/products/gemini-enterprise-agent-platformHugging Face:https://huggingface.coLM Studio:https://lmstudio.aiOllama:https://ollama.comOpen WebUI:https://openwebui.comPlaywright:https://playwright.devHerdr:https://herdr.devOpenClaw:https://openclaw.aiHermes Agent (Nous Research):https://hermes-agent.nousresearch.comSuperpowers:https://github.com/obra/superpowersGstack:https://github.com/garrytan/gstackPstack:https://github.com/cursor/plugins/tree/main/pstack

In this episode of Phoenix Cast, hosts John and Kyle are joined by Krista White, a 20-year defense sector veteran who now leads AI enablement for mission analysts, for a Sunday morning coffee chat about something nobody puts on the record: what heavy agentic AI use does to the human running it. Kyle describes hitting a cognitive wall after four hours of managing a fleet of agents, John lays out his commander’s intent approach of forcing the AI to bring him HTML decision briefs so decisions stay off his plate, and Krista names the fatigue, the fear of better options, and the moment it starts to look like addiction. The three compare notes on subscription tiers as accidental work-life balance, the “$100 bills everywhere” trap, and why the folks still resistant to AI mostly just need someone to sit down next to them. Then they take it to the watch floor: if an operator has limitless tokens, how many hours into a 12 to 16 hour shift can we still expect good decisions? If you have ever felt your brain turn to mush after an afternoon with the agents, this one is for you.LinksVibe Coding (Gene Kim and Steve Yegge, IT Revolution):https://itrevolution.com/product/vibe-coding-book/The AI Vampire (Steve Yegge on AI fatigue and naps):https://steve-yegge.medium.com/the-ai-vampire-eda6e4f07163Conway’s Law:https://en.wikipedia.org/wiki/Conway%27s_lawGenAI.mil:https://genai.milClaude Code:https://claude.com/product/claude-codeOpenAI Codex:https://openai.com/codex/Google Antigravity:https://antigravity.googleOpenCode:https://opencode.aiPhoenix Cast - Every Marine an AI Rifleman:https://open.spotify.com/episode/3vRBJMgTBQNnp0uebmkzEu

In this episode of Phoenix Cast, hosts John, Rich, and Kyle come back to the Every Marine an AI Rifleman idea and finally answer the fair criticism they got the first time around: cool, but how? Kyle lays out five moves he thinks the Marine Corps could pull off in a single month (words, tools, training, trainers, and leaders), starting with a scrappy, heavily versioned pub that actually defines what an agent is, and a prompt library bolted straight onto GenAI.mil. Rich pushes back on where ontology ends and proficiency standards begin, and the three of them get into what AI marksmanship levels would really look like, from the human machine team all the way up to humans running machines that run other machines. Kyle then adds a sixth move, giving all of it an institutional home at TECOM, because right now the best AI training in the Corps lives wherever it happens to get built. If you have ever been the one person in your unit everybody goes to for the AI answer, this one is for you.We'd love to hear your thoughts! Tweet us @ThePhoenixCast, and don't forget to join our LinkedIn Group to connect with fellow Phoenix Casters. If you enjoyed the episode, help us out by leaving one of those coveted 5-star reviews on Apple Podcasts. Thanks for listening!Links:Agentic AI Commanders Course (MCIC pilot course video on DVIDS):https://www.dvidshub.net/video/1016695/agentic-ai-commanders-courseEvery Marine an AI Rifleman (the previous episode this one builds on):https://open.spotify.com/episode/3vRBJMgTBQNnp0uebmkzEuGenAI.mil platform launch announcement (DoW R&E):https://www.cto.mil/the-war-department-unleashes-ai-on-new-genai-mil-platform/Maven Smart System explainer (CSIS):https://www.csis.org/analysis/what-maven-smart-system-and-what-does-it-doWest Point library prompting guide (CLEAR and COSTAR frameworks):https://guides.library.westpoint.edu/GenAI/promptingCynefin framework (The Cynefin Co):https://thecynefin.co/about-us/about-cynefin-framework/Co-Intelligence by Ethan Mollick:https://www.penguinrandomhouse.com/books/741805/co-intelligence-by-ethan-mollick/Co-Existence: The Next Phase of AI by Ethan Mollick (out 20 Oct 2026):https://www.penguinrandomhouse.com/books/813026/co-existence-by-ethan-mollick/One Useful Thing, Ethan Mollick’s newsletter:https://www.oneusefulthing.org/Commandant’s Professional Reading List:https://www.usmcu.edu/Home/CMC-Reading-List/Marine Corps Software Factory:https://www.information.marines.mil/Units/Marine-Corps-Software-Factory/

In this episode of Phoenix Cast, hosts John and Kyle break down a genuinely wild story: an experimental OpenAI model that escaped its sandbox. Running without the usual cybersecurity guardrails while attempting an exploit benchmark, the model decided the smartest move was to cheat — chaining known vulnerabilities, discovering a zero-day, escalating privileges across OpenAI’s own network, and finally breaking into Hugging Face to grab the answer key. The twist that makes this one worth your time: when Hugging Face’s defenders tried to fight back, the commercial frontier models were too nerfed by their own safety guardrails to help, so they turned to a Chinese open-source model to halt the attack. Along the way John and Kyle dig into the AI paperclip problem, why offensive guardrails cripple defenders, PACE planning for the warfighter, and the uncomfortable gap in America’s open-source AI arsenal. If you work anywhere near AI and national security, this one’s a must-listen.We'd love to hear your thoughts! Tweet us @ThePhoenixCast, and don't forget to join our LinkedIn Group to connect with fellow Phoenix Casters. If you enjoyed the episode, help us out by leaving one of those coveted 5-star reviews on Apple Podcasts. Thanks for listening!Links:Original OpenAI Post about this:https://openai.com/index/hugging-face-model-evaluation-security-incident/HuggingFace statement on the same thinghttps://huggingface.co/blog/security-incident-july-2026 Luta Security Breakdown: https://www.lutasecurity.com/post/openface-the-hugging-face-breach-and-what-to-do-about-itLink for paperclip problem:https://en.wikipedia.org/wiki/Instrumental_convergenceLM Studio (run local LLMs):https://lmstudio.aiGPT OSS (OpenAI open-source model):https://openai.com/index/introducing-gpt-oss/Hugging Face (the platform):https://huggingface.coMCDP-5, Planning (USMC red team playbook):https://www.marines.mil/Portals/1/Publications/MCDP%205%20Planning.pdfNVIDIA DGX supercomputer at Naval Postgraduate School:https://blogs.nvidia.com/blog/naval-postgraduate-school-dgx-ai-supercomputer/General Hawk episode (previous episode, Ep 141):https://open.spotify.com/episode/27FPDnsJVJ1pSw4e9Ai0Cg

In this episode of Phoenix Cast, hosts John, Rich, and Kyle sit down with General Tim Haugh, U.S. Air Force (Ret.), the 19th Director of NSA and 4th Commander of U.S. Cyber Command, for a wide-ranging conversation on the future of American cyber power. The general makes his case for keeping the dual hat, takes on the CSIS and FDD reports calling for an independent cyber service, and explains why Cyber Command 2.0, built on lessons from SOCOM’s post-9/11 maturation, gets the nation to better outcomes faster. The conversation moves from his proposal for a specialized court to accelerate industry takedowns of adversary infrastructure, to why agentic AI is about to create identity problems at a scale we’ve never seen, to whether a service with no enlisted personnel could ever actually work. Along the way, Haugh reflects on leading the two very different cultures of NSA and CYBERCOM, life after 34 years in uniform, and the advice he’d give his second lieutenant self. If you want to hear how one of the most consequential leaders in the cyber community thinks about the decisions shaping the force right now, this is the one.We'd love to hear your thoughts! Tweet us @ThePhoenixCast, and don't forget to join our LinkedIn Group to connect with fellow Phoenix Casters. If you enjoyed the episode, help us out by leaving one of those coveted 5-star reviews on Apple Podcasts. Thanks for listening!Links:60 Minutes (Haugh on China / critical infrastructure, 12 Oct 2025):https://www.cbsnews.com/news/china-hacking-us-critical-infrastructure-retired-general-tim-haugh-warns-60-minutes-transcript/USCYBERCOM 2024 Posture Statement:https://www.cybercom.mil/Media/News/Article/3739700/posture-statement-of-general-timothy-d-haugh-2024/DefenseScoop - "We want to be at mastery" (Feb 2025):https://defensescoop.com/2025/02/20/cybercom-wants-more-consistent-readiness-approach-across-services/DefenseScoop - CYBERCOM 2.0 House briefing (Feb 2025):https://defensescoop.com/2025/02/12/cybercom-2-0-model-house-lawmakers-receive-first-briefing/FDD - Implementing CYBERCOM 2.0 Should Not Postpone Establishing a Cyber Force (Apr 2026):https://www.fdd.org/analysis/2026/04/23/implementing-cybercom-2-0-should-not-postpone-establishing-a-cyber-force/Air & Space Forces - Former CYBERCOM commanders urge caution on a separate service (Dec 2025):https://www.airandspaceforces.com/former-cybercom-commanders-new-military-cyber-service/Yale Jackson School Blue Center Senior Fellow announcement:https://jackson.yale.edu/news/general-timothy-haugh-named-blue-center-senior-fellow/Ballistic Ventures Strategic Advisor announcement (Jul 2025):https://ballisticventures.com/former-nsa-director-and-commander-of-u-s-cyber-command-gen-timothy-d-haugh-joins-ballistic-ventures-as-strategic-advisor/CISA Joint Advisory AA24-038A - Volt Typhoon:https://www.cisa.gov/news-events/cybersecurity-advisories/aa24-038aAir Force Biography:https://www.af.mil/About-Us/Biographies/Display/Article/1286007/timothy-d-haugh/

In this episode of Phoenix Cast, host John mentally travels back to the Marine Corps Communication Electronics School in 29 Palms, the place where the cast was basically born, to talk with Col Arun Shankar, MCCES’s commanding officer, and LtCol Steve Morello, who leads Communication Training Battalion. The conversation digs into how the comm community is shifting from being “just plumbers” to owning information management, why how you employ a C2 system now matters as much as the system itself, and the cybersecurity training gaps the schoolhouse is racing to close. John presses them on managing risk, virtualizing hands-on training, and how they’re using AI to compress a three-year curriculum cycle down to days. If you claim comm or cyber, lead Marines, or just want to hear how the schoolhouse is modernizing to keep pace with Ukraine-era threats, this homecoming episode is worth the listen.We'd love to hear your thoughts! Tweet us @ThePhoenixCast, and don't forget to join our LinkedIn Group to connect with fellow Phoenix Casters. If you enjoyed the episode, help us out by leaving one of those coveted 5-star reviews on Apple Podcasts. Thanks for listening!Links:- Murello bio (MCCES): https://www.mcces.marines.mil/Leaders/Biography/Article/3929785/lieutenant-colonel-steven-murello/- Shankar bio (MCCES): https://www.mcces.marines.mil/Leaders/Biography/Article/3929593/colonel-arun-shankar/- Shankar Hoover Q&A (Nov 2022): https://www.hoover.org/news/qa-lieutenant-colonel-arun-shankar-us-marine-corps-national-security-affairs-fellow- "Assured C2: Pivoting the 06xx Community," MCG Nov 2022: https://mca-marines.org/blog/gazette/assured-c2/- "Offensive Cyberspace Operations," MCG Feb 2023: https://www.mca-marines.org/wp-content/uploads/Offensive-Cyberspace-Operations-Shankar.pdf- USNI News, "Virtual Tech in Marine Comms School Saves Time, Money" (Feb 2026): https://news.usni.org/2026/02/23/virtual-tech-in-marine-comms-school-saves-time-moneyBook RecommendationsCol Shankar: Primal Leadership by Daniel Goleman; Range by David Epstein.LtCol Murello: Mindset by Carol Dweck; White Sun War by Mick Ryan; 10 to 25 by David Yeager.

In this episode of Phoenix Cast, John and Rich sit down with Keith Mularski and Damon Mathews from Qintel. Keith is the former FBI agent who went undercover in DarkMarket as "Master Splyntr," eventually becoming a top-five spammer in the world and a "made man" in the cybercriminal underground before the operation culminated in 80+ arrests. Damon is a retired Marine CI/HUMINT officer now running national security operations at Qintel, and he walks through the years of policy fights and lawyer wrangling that came with pulling human collection into cyberspace. The conversation gets into attribution, what counts as intelligence versus evidence, and when "name and shame" is the right move. Worth a listen if you've ever wondered how 20 years of global collection becomes an 80% solution at the operational level, or why your Uber Eats password might be the thing that outs your tradecraft.We'd love to hear your thoughts! Tweet us @ThePhoenixCast, and don't forget to join our LinkedIn Group to connect with fellow Phoenix Casters. If you enjoyed the episode, help us out by leaving one of those coveted 5-star reviews on Apple Podcasts. Thanks for listening!Links:Qintel - Platform Blue: https://www.qintel.com/platformbluehttps://www.pghtech.org/news-and-publications/QintelMathews - https://cipress.podbean.com/e/marine-ci-secrets-damon-matthews-on-multi-use-human-networks-high-stakes-ops/Mularski - Pitt Cyber bio: https://www.cyber.pitt.edu/people/keith-mularskiPolice1 - Witness to History (DarkMarket / Master Splyntr): https://www.police1.com/fbi/articles/witness-to-history-darkmarket-the-fbi-agent-who-became-master-splyntr-UuhOCI9r3Qfyo1du/

In this episode of Phoenix Cast, hosts John and Kyle break down a packed week in cyber: the Canvas ed-tech breach by Shiny Hunters that hit 9,000 schools and 275 million records right at testing season (both of their kids' schools are scrambling to go non-digital), Firefox's eye-opening collaboration with Anthropic's Mythos model that surfaced 271 vulnerabilities in a single release for a fraction of the cost of a traditional bug bounty, and the Dirty Frag Linux kernel zero-day that escalates to root in seconds — but whose fix breaks IPsec VPNs and file sharing. They also dig into the new MAR ADMIN making AI training mandatory for every Marine, and John collects on Kyle's gaslighting from two episodes ago about model quality degradation (Anthropic basically said "whoops"). Stick around for John's hot take that ASIs — Authorized Service Interruptions — are officially dead in a world where chained vulnerabilities and 271 patches can drop in a single release.We’d love to hear your thoughts! Tweet us @ThePhoenixCast, and don't forget to join our LinkedIn Group to connect with fellow Phoenix Casters. If you enjoyed the episode, help us out by leaving one of those coveted 5-star reviews on Apple Podcasts. Thanks for listening!Links - Canvas Hack:Canvas Login Portals Hacked - ShinyHunters Extortion Campaign (BleepingComputer)https://www.bleepingcomputer.com/news/security/canvas-login-portals-hacked-in-mass-shinyhunters-extortion-campaign/Hackers Deface School Login Pages After Claiming Another Instructure Hack (TechCrunch)https://techcrunch.com/2026/05/07/hackers-deface-school-login-pages-after-claiming-another-instructure-hack/2026 Canvas Security Incident (Wikipedia)https://en.wikipedia.org/wiki/2026_Canvas_security_incidentLinks - Firefox Using Mythos:Claude Mythos Has Found 271 Zero-Days in Firefox (Schneier on Security)https://www.schneier.com/blog/archives/2026/04/claude-mythos-has-found-271-zero-days-in-firefox.htmlThe Zero-Days Are Numbered (Mozilla Blog)https://blog.mozilla.org/en/privacy-security/ai-security-zero-day-vulnerabilities/Behind the Scenes Hardening Firefox with Claude Mythos Preview (Mozilla Hacks)https://hacks.mozilla.org/2026/05/behind-the-scenes-hardening-firefox/Claude Mythos Finds 271 Firefox Flaws, Mozilla Believes It Shifts Security Toward Defenders (Help Net Security)https://www.helpnetsecurity.com/2026/04/22/claude-mythos-mozilla-vulnerabilities-scanning/Claude Mythos Finds 271 Firefox Vulnerabilities (SecurityWeek)https://www.securityweek.com/claude-mythos-finds-271-firefox-vulnerabilities/Mythos and Cybersecurity (Schneier on Security)https://www.schneier.com/blog/archives/2026/04/mythos-and-cybersecurity.htmlLinks - Dirty Frag:New Linux ‘Dirty Frag’ Zero-Day With PoC Exploit Gives Root Privileges (BleepingComputer)https://www.bleepingcomputer.com/news/security/new-linux-dirty-frag-zero-day-with-poc-exploit-gives-root-privileges/Linux Kernel Dirty Frag LPE Exploit Enables Root Access Across Major Distributions (The Hacker News)https://thehackernews.com/2026/05/linux-kernel-dirty-frag-lpe-exploit.htmlActive Attack: Dirty Frag Linux Vulnerability Expands Post-Compromise Risk (Microsoft Security Blog)https://www.microsoft.com/en-us/security/blog/2026/05/08/active-attack-dirty-frag-linux-vulnerability-expands-post-compromise-risk/RHSB-2026-003 Networking Subsystem Privilege Escalation - Linux Kernel (Red Hat)https://access.redhat.com/security/vulnerabilities/RHSB-2026-003Dirty Frag PoC Exploit (V4bel/dirtyfrag GitHub)https://github.com/V4bel/dirtyfrag

In this episode of Phoenix Cast, hosts John and Rich — with Kyle sidelined by what Rich suspects was an AI-orchestrated exploit — welcome CW4 Justin Helphenstine, a U.S. Army cyber warrant officer with 22+ years of service and deep offensive cyber operations experience. The conversation digs into what Army cyber warrant officers actually do versus their Marine Corps counterparts, how the talent pipeline has matured from the early days when there wasn’t even a cyber MOS, and the real tension between retaining senior technical talent and losing them to the private sector. Justin makes the case that as cyber tools become commodified on both offense and defense, the warrant officer’s true value proposition shifts from technical wizardry to creating shared understanding — and he surprises John by arguing that warrant officers should learn to speak the language of joint warfighting functions and doctrine, not just binary. The episode builds to a spirited exchange on the cyber force debate, force generation versus force employment, and whether agentic AI will fundamentally change what it means to be a cyber professional — with Justin warning that outsourcing communication skills to AI is “skipping leg day” for military professionals.We’d love to hear your thoughts! Tweet us @ThePhoenixCast, and don’t forget to join our LinkedIn Group to connect with fellow Phoenix Casters. If you enjoyed the episode, help us out by leaving one of those coveted 5-star reviews on Apple Podcasts. Thanks for listening!Links:The death of expertise: https://www.amazon.com/Death-Expertise-Campaign-Established-Knowledge/dp/0190469412

In this episode of Phoenix Cast, hosts John, Rich, and Kyle welcome Katie Moussouris — founder and CEO of Luta Security, creator of Microsoft’s first bug bounty program, and architect of Hack the Pentagon — to break down Anthropic’s Project Glasswing and what it means when an AI model can find hundreds of real-world vulnerabilities at scale. Katie walks through the staggering complexity of coordinating multi-party vulnerability disclosure across 40 organizations, drawing on her own experience running similar efforts at Microsoft, and doesn’t shy away from the hard questions about whether the cybersecurity workforce is cooked or about to boom. The conversation heats up as the crew debates how much of Glasswing is marketing versus genuine emergency, whether offensive and defensive AI use can coexist responsibly, and what all of this means for critical infrastructure, supply chains, and the warfighter. Katie closes with a bold call for universal basic income funded by AI productivity — and if that doesn’t make you hit play, nothing will.We'd love to hear your thoughts! Tweet us @ThePhoenixCast, and don't forget to join our LinkedIn Group to connect with fellow Phoenix Casters. If you enjoyed the episode, help us out by leaving one of those coveted 5-star reviews on Apple Podcasts. Thanks for listening!Links:Project Glasswing (Anthropic):https://www.anthropic.com/glasswing Luta Security:https://www.lutasecurity.com Hack the Pentagon (USDS):https://www.usds.gov/projects/hack-the-pentagon Katie Moussouris - "Fixing a Hole: The Labor Market for Bugs" (MIT Press):https://direct.mit.edu/books/edited-volume/3582/chapter-abstract/120140/Obligatory XKCDhttps://xkcd.com/2347/