The Models Can Hurt You

The Models Can Hurt You

After a two-month hiatus, Barrett and Greg return to Vibe Coping with a more personal opening than usual. Barrett talks about the unexpected passing of his sister Erika, the strange way grief can make tech discourse feel trivial, and why returning to conversation itself feels meaningful. From there, the episode moves into the rapidly shifting world of AI: Greg reports from Bali, where Claude Code sightings are apparently as common as flat whites, and the two discuss whether code is still a moat, what happens when everyone can build a SaaS app, and whether every business may eventually need an “AI administrator” to manage its internal tools. They dig into the release and sudden withdrawal of Anthropic’s Fable Five model, the experience of using more powerful coding agents, and the unsettling feeling of becoming dependent on tools that can disappear overnight. They also talk about the limits of AI-generated work: the final-mile problem, generic content, security risks, leaked keys, AI-assisted codebases no one fully understands, and the emerging backlash against AI-generated contributions in certain programming communities. The conversation widens into moats, inherited assumptions, trillionaire Elon Musk, SpaceX, low Earth orbit, satellite warfare, GPS disruption, Russia’s role in cyber and AI, and the broader race to control both models and infrastructure. Later, Barrett and Greg unpack the phrase “no prompts, just loops,” and what it says about the future of agentic work: defining a goal, letting the system iterate, and building verification into the loop. Barrett shares a more optimistic personal angle, describing how AI helped him process complex medical information during his sister’s health crisis and gave him a sense of relief around long-delayed technical projects. Greg counters with the anxiety, guilt, and broken flow state of working alongside agents. Finally, they touch on Zcash, privacy coins, the risks of cryptographic bugs, and the migration of crypto hype into AI consulting. Topics include grief, AI coding agents, Claude Code, Fable Five, software moats, vibe coding, security risks, handwritten code, SpaceX, Elon Musk’s wealth, satellite warfare, Russia, agent loops, productivity, parenting, gardening, Zcash, crypto, and the strange emotional texture of trying to live through technological acceleration.

"The Dumbest Person You Know Is Up 40%"

* 00:00 — Cold Open: "The Casino" (1920s/2020s parallels) * 04:30 — What are prediction markets? Kalshi, Polymarket, and the $44B industry * 07:30 — The Iran insider trading scandal ($553K on a single bet) * 11:00 — Where's the line between information and insider trading? * 15:00 — Financial nihilism: why young people are gambling instead of saving * 19:00 — Greg's GMG penny stock saga * 26:30 — AI data centers: how bad are they, really? * 33:00 — National security vs. community opposition * 36:00 — The Claude Mythos leak * 44:00 — 20-year-old Linux bugs and the cybersecurity arms race * 49:00 — Claude Code source leak and AI IP law * 52:00 — "People aren't ready for what's coming" — but what does that mean? * 54:00 — Digital identity, deepfakes, and family safety passcodes * 58:00 — Greg tests his bank's security and locks himself out of his credit card * [BONUS] — Interview with Claude Mythos

Dispatches from the Belly of the AI Beast (feat. The Blue Power Ranger)

In Session 3 of Vibe Coping, Barrett and Greg sit down with their very first guest, Billy! Broadcasting from the heart of the Bay Area, Billy takes us through his journey from a Vancouver coding bootcamp to building JobGet, a platform serving over 100 million hourly job seekers. The guys dive deep into the reality of deploying AI in a massive codebase, the power of Claude’s sub-agent teams, and what the AI and robotics scene really looks like in China. Plus, we debate Anthropic's labor market report to figure out if software will eat the world, or if we'll just end up needing more plumbers. Key Takeaways * Meet Billy & The Origins of JobGet: Billy shares his background, from growing up in China and Vancouver to moving to the Bay Area. He discusses meeting Barrett at a 2014 coding bootcamp and eventually scaling JobGet, an hourly job marketplace that has raised over $60 million. * Blue-Collar Work in a Tech World: Why JobGet focuses specifically on hourly workers instead of the white-collar roles usually targeted by tech platforms. The impact of the pandemic and changing economic conditions on industries like healthcare and trucking. * The Reality of AI in Engineering: How Billy’s team uses Claude Code internally, particularly for navigating massive repositories acquired from companies like Snagajob. The reality of vibe coding, dealing with AI regressions, and refactoring codebases within 30 days. * Claude Teams & Resurrecting Old Apps: Billy explains how he used Claude sub-teams—assigning AI personas like a PM, engineer, and VC—to successfully rebuild a 10-year-old app idea (MIT) over the weekend. * Prompting: Skill Issue or Trust Issue?: The guys debate whether bad AI outputs are a result of poor prompting skills, a lack of trust in the tool, or simply a lack of foundational system design knowledge. * The Tech Scene in China: Billy shares observations from his recent trips to China, detailing how the population is being taught to use Open-Claude and the massive leap forward in robotics showcased at the Spring Gala. * The SF AI Bubble & The Future of Jobs: Discussing the AI-obsessed billboard culture of San Francisco. The hosts analyze Anthropic's labor market impact report, debating which jobs are truly "AI-proof" and why physical-world complexity (like plumbing) might keep humans employed. * Fun Fact: Billy reveals the origin of his name: he named himself after the Blue Power Ranger! Timestamps * 00:00 — Billy's intro (the full roast) * 01:02 — Billy's background: Xi'an, Surrey, Vancouver coding bootcamp * 07:13 — What is JobGet? The hourly labor marketplace * 09:20 — How AI and the post-pandemic economy have hit the hiring market * 12:19 — Greg asks about VC expectations in the AI era * 18:00 — Dev tools and workflow: Claude Code, sub-teams, and what engineering leaders are actually using * 25:47 — Vibe coding gone wrong: the LLM gateway that needed a full refactor * 30:05 — Claude sub-teams experiment: spinning up agent teams with roles * 31:05 — Resurrecting MIT: the travel/events app Barrett and Billy built 10 years ago * 35:54 — "AI is a multiplier" — skill issue vs. trust issue debate * 42:00 — OpenClaw, Claude Dispatch, and the tool churn cycle * 44:33 — China's push to democratize AI: OpenClaw workshops and the Spring Gala robots * 49:40 — US-China AI rivalry, TikTok, Chinese EVs, and propaganda * 56:57 — San Francisco as AI ground zero: billboards, coffee shops, and being numb to it * 59:58 — "Software is eating the world" — did it actually? Will AI? * 1:05:15 — Anthropic's labor market report and what's left for humans * 1:11:22 — Embodied AI, world models, and robot economics * 1:12:42 — What's AI-proof: sports, cults, and handmade goods * 1:13:31 — Why Billy is named Billy (the Blue Power Ranger)

Prompt the DJ, Simulation & Surrender

The rebrand is official — Terminal is dead, Vibe Coping is alive. We jump into AI-generated music, debating whether fake artists can build real fanbases and floating the idea of a DJ who prompts AI in real time to read the crowd. From there it's AGI, sycophancy, and trust: why handing your decisions to ChatGPT is basically the tech version of "Jesus take the wheel," and why that's sketchy when the models still confidently make stuff up. They trade stories about AI giving panic-level medical advice over minor cuts. Then Greg goes full galaxy brain, laying out why the universe looks like an optimized video game — the observer effect as lazy rendering, the speed of light as a refresh rate, black holes as computational bottlenecks. 0:00 — Intro & the rebrand from Terminal to Vibe Coping3:07 — AI music artists: Velvet Sundown, Zanya Monet, BBL Drizzy5:35 — Do artists need to be compelling people to break through?9:13 — "Prompt the DJ" — could prompting be a new art form?11:02 — AI slop, influencers & the emotional connection question14:38 — Two-way content: AI music that adapts to your heart rate16:24 — Shared experiences vs. personalized content19:05 — Cloudflare, Anthropic's C compiler, and where the real value in code lives25:34 — The bullshit meter test: ChatGPT vs. Claude vs. Gemini27:45 — Are we hitting the hockey stick on the AI intelligence curve?30:08 — AGI debate: can AI solve problems without human hand-holding?33:17 — "Jesus take the wheel" — trusting AI like trusting God40:42 — Simulation theory: does living through the singularity feel simulationy?43:04 — Greg's randomness obsession & the double slit experiment46:40 — The universe as a video game: render loops, jank & black holes48:58 — View layer vs. data layer: quantum entanglement as buffer overflow52:18 — Anthropic vs. the US military & the autonomous weapons red line57:21 — Will governments seize AI? Leopold Aschenbrenner's thesis1:05:23 — Sam Altman's "raising a human" comment & the Linux analogy1:08:28 — Wrap up

Session 1

Summary In this episode of the Terminal podcast, Barrett and Greg explore the rapidly changing landscape of technology and finance, focusing on the implications of AI advancements, the challenges faced by new graduates, and the evolving nature of software development. They discuss the potential of AI tools like OpenClaw, the impact of uncertainty on career paths, and the future of personal computing. The conversation also touches on investment strategies in the software sector and the importance of understanding the context in which AI operates. Takeaways * The first podcast was a Grateful Dead song. * Navigating the AI revolution presents both challenges and opportunities. * Uncertainty in the job market can create unique opportunities for innovation. * AI tools are evolving, but their utility is still being explored. * The future of personal computing may shift away from traditional screens. * OpenClaw represents a new frontier in AI interaction. * AI's role in relationships raises questions about decision-making. * The hype cycle around AI tools can lead to confusion and misalignment. * Investing in software companies requires understanding their long-term viability. * The integration of AI into existing systems will shape the future of work.