RECENT EPISODES
Mon 21 SeptThe Right Way To Write With AIIn this episode, Nicholas Cole explores the evolving landscape of writing in the age of AI, emphasizing the need for engaging, original content that resonates with audiences. He argues that while AI can automate basic writing tasks, it cannot replicate personal voice, making unique storytelling essential for standing out.Fri 18 SeptJev is HERE. How to use itIn this episode, we explore Jev, a groundbreaking AI technology that excels in classifying emails with remarkable speed and affordability. Unlike traditional large language models, Jev categorizes emails into segments like shopping and finance, offering a spam score and urgency assessment. With processing times around 200 milliseconds and costs as low as 18 cents for 1,700 emails, Jev presents a cost-effective solution for businesses, though users are advised to employ it cautiously in advisory roles rather than for critical decisions.Tue 15 SeptInstinct AI: The AI Assistant for normal peopleIn this episode, we explore Instinct AI, a $10 billion invite-only personal assistant app designed for everyday users, including those less familiar with technology. The app excels in managing tasks like booking flights and ordering food while utilizing Apple iMessage for human-like interactions. Despite some limitations, such as not messaging WhatsApp directly, the speaker highlights its user-friendly onboarding and time-saving potential, while also addressing privacy concerns that lead some users to prefer established tech companies.Mon 14 SeptBuilding a Software Factory that actually works (Full Course)In this episode, listeners will learn how to build an effective software factory that enhances software development efficiency through AI agents, akin to an assembly line. The speaker emphasizes the importance of understanding workflow processes and domain knowledge to optimize model capabilities while maintaining quality. Key topics include isolating new features to prevent conflicts, the significance of evidence-driven testing, and the use of code review agents to ensure high standards before merging code into the main application.Thu 10 SeptYou're using GPT-6 Astra WRONGIn this episode, Ross Mike reveals how users are underutilizing GPT-6 Astra, particularly in launching hardware startups and automating tasks. He emphasizes the importance of performance updates and security audits for applications using the tool, while also advocating for the use of reference images in UI design.Tue 8 SeptLocal AI Clearly ExplainedIn this episode, the hosts delve into the transformative potential of Local AI, which operates on user-controlled hardware, offering enhanced privacy and offline functionality. They explore key components of the local AI landscape, including model warehouses like Huggingface and user-friendly interfaces such as LM Studio. The discussion highlights quantization techniques for model compression and the practical applications of Local AI in niche businesses managing sensitive data, emphasizing the importance of starting with simple tasks to maximize efficiency.Wed 2 SeptThese 5 Github Repos are a goldmineIn this episode, listeners discover five invaluable GitHub repositories that can significantly boost product development and efficiency in the evolving tech landscape. The discussion highlights Peter Yang's no AI slop skill, which enhances AI-generated writing, and a CRM by Tri-Comp AI that optimizes relationship management for businesses. Additionally, the episode covers tools for video editing and security scanning of AI skills, emphasizing the importance of exploring GitHub to stay ahead in innovation and productivity.Mon 31 AugMaking $$$ as a Marketing EngineerIn this episode, we explore the emerging role of marketing engineers, projected to earn between $250,000 and $1 million as they leverage AI to automate marketing tasks. The discussion highlights the importance of converting market signals into actionable sales strategies and establishing a growth repository to enhance marketing efforts. We also delve into the necessity of regular tool analysis and the integration of diverse skills to effectively utilize AI for driving startup growth.Wed 26 AugWebMCP clearly explained (and how to make $$)In this episode, we explore WebMCP, a groundbreaking feature from Google and Microsoft that enhances AI agent interactions with websites, streamlining user engagement. As this technology gains traction in Chrome, it presents significant opportunities for businesses to improve customer experiences and drive sales, particularly in complex purchasing sectors. We also discuss how startups can capitalize on WebMCP by offering setup services and user journey reports, creating new revenue streams in the evolving digital landscape.Tue 25 AugScreensharing top takes in AI/startupsIn this episode, Greg Eisenberg discusses the essential steps for making a company AI native, emphasizing the need for clear documentation and gradual integration of AI tools. The conversation highlights the success of the baby food app "Solid Starts," illustrating a trend toward premium offerings in niche markets. Additionally, the potential of iMessage agents and the evolving role of customer support in product development are explored, underscoring the importance of balancing technology with human interaction in business.Fri 21 AugGrok Bot: make a 1 person company with agentsBilly Howell makes the case that Grok Bot, priced between $200 and $300 per month, is the strongest available platform for running agent teams and that a single person can use it to operate a real business, as he does with his Arlington Bagel newsletter serving 6,000 subscribers.Wed 19 AugYou need to be skillsmaxxing (10x your Claude/Codex)Remy makes the case that agent skills, defined as SOPs packaged as markdown files that Claude reads to execute a process consistently, are the highest-leverage thing someone can build with AI right now, estimating a single well-built skill saves two hours a week with compounding returns as more are stacked. He walks through his thin agents, thick skills philosophy and specific examples including proposal generation, brand voice, and a YouTube publish workflow structured as a skill chain.Mon 17 AugHow to use Claude Code better than 99% of PeopleMost users treat Claude Code as a chat interface and get mediocre results, but the episode argues that configuring it as a structured AI employee across nine defined areas is what separates effective use from wasted potential.Wed 12 AugHow to Build an AI-Native Company in 2026Ali Miller, who once managed roughly 100 people at AWS, now runs her company with 34 AI agents organized under six director-level roles, including a chief dreaming officer whose sole job is asking how to 10x any output. She explains that models like GPT 5.6 now respond meaningfully to vague prompts such as her three-word favorite, do smart things, a capability she says did not exist a year ago.Mon 10 AugMaking $$$ selling to AI AgentsCloudflare's new pay-per-crawl system, built on the HTTP 402 status code and a payment rail called X402, lets site owners charge AI agents directly at the edge before a request reaches the origin server, replacing the broken bargain where crawlers traded access for human traffic.Wed 5 AugThese AI Marketing Agents Get You CustomersCody Schneider argues that marketing agents are the direct successor to coding agents, making customer acquisition on demand the same way coding agents made software development on demand.Mon 3 AugGraph Engineering Clearly ExplainedGraph engineering, defined as designing how work moves between steps in an AI system, solves a core problem with single-prompt AI: one model is forced to research, interpret, write, and grade its own confidence simultaneously, producing unchecked overconfidence. The solution is a graph structure where parallel nodes handle separate tasks, a skeptic node checks claims, and a human gate precedes expensive decisions.Tue 28 JulJack Dorsey's Buzz: The New Hermes Agent?Jack Dorsey's Buzz is a team collaboration tool built on the open Nostr protocol that positions itself as a competitor to both Slack and GitHub by storing all chats, agent interactions, and data on user-controlled relays rather than locking data inside a proprietary platform. Agents are first-class team members in Buzz, with a swappable harness supporting Claude Code, Codex, and Goose without losing conversation history, and tight Git integration lets agents spin up parallel work trees simultaneously.Mon 27 JulHow I use Claude Code + MCPs to run my marketingCody Schneider draws a hard line between genuine marketing agents and the linear automation workflows most builders are actually shipping in tools like Zapier or N8n. A real agent, he argues, requires a unified data pipeline connecting Facebook ads, Google Analytics, HubSpot, and Stripe into a warehouse like ClickHouse so the system can tie specific creative to revenue, then autonomously run a thinking loop on a recurring cadence.Fri 24 JulMost Valuable Skill of 2026: Managing AI AgentsRyan Carson, the sole employee of AI divorce startup Untangle, makes the case that managing teams of AI agents will be the most valuable skill for founders and solopreneurs in 2026, drawing on his experience scaling Treehouse to 110 employees as a direct point of comparison. Carson runs five to ten cloud-based agents simultaneously through Devin, shipping up to 40 pull requests per day, and warns that working locally produces roughly ten times less output.Fri 24 JulHow I run a team of AI Agents 24/7Ryan Carson, solo founder of Untangle, an AI divorce agent serving family law firms, explains how running five to ten parallel cloud-based AI agents simultaneously allows him to ship an average of 22 to 25 pull requests per day. He argues that local development cannot scale to multi-agent workflows because code copies collide, making cloud environments from tools like Devin, Codex, and Claude Code essential.Mon 20 JulFDE: The $1M/Year AI Job ExplainedThe forward-deployed engineer, or FDE, is the role Palantir popularized by embedding engineers directly at enterprise and government clients to map real workflows and build targeted AI solutions, with compensation reaching one million dollars per year for top candidates. The core reason the role commands that pay is that frontier AI intelligence has been commoditized, shifting competitive advantage to where and how it is deployed.Wed 15 JulThe $1,000/hour Solo AI business (Full Course)Corey Gannon built a solo consulting business around a $999 AI tools assessment targeting small businesses with two to twenty employees and $500,000 to $5 million in revenue, requiring no coding, no audience, and no capital. The workflow uses Claude to analyze discovery call transcripts, match pain points to off-the-shelf tools from directories like Futurepedia.io and There's An AI For That, and generate a client-facing report in Claude Artifacts.Mon 13 JulMaking $$$ with Loop EngineeringLoop engineering, a build-verify-iterate cycle for AI agents drawing on lean startup principles, is the central focus of this episode. Ellie walks through his production SEO loops for Inbox Zero and draftfantasy.com, where agents pull real ranking data from Google Search Console and the Data for SEO API, make page-level changes, and run every two weeks to one month for under five dollars in API costs.Fri 10 JulGrok 4.5 is a bigger deal than FableNick from Orgo argues that Grok 4.5 is the most significant model upgrade his Hermes and OpenClaude agent stack has seen, offering intelligence comparable to Anthropic Opus 4.8 at roughly one tenth the cost and ten to fifteen times the speed of Fable for equivalent tasks. In a live demo, his Dewey agent completed a full landing page build in about 40 seconds while a simultaneously running GPT 5.6 agent was still processing.Thu 9 JulWe Tested OpenAI's GPT 5.6 for a MonthAfter a month of daily use across email, writing, and model training, Dan Chipper rates GPT 5.6 as the most usable and fastest model available for knowledge work, placing it at roughly 60 to 70 percent of Claude Fable's capability while noting Fable remains too slow and expensive for most practical tasks.Wed 1 JulAI Agents are the new SaaSAI agents are framed as the new SaaS because they shift from helping users do work to doing the work itself, competing for human labor budgets rather than software tooling spend, which makes the addressable market far larger than traditional SaaS. The episode walks through what makes an agent commercially viable, citing Slang AI for restaurant call handling and Same Day for home services dispatching as examples of the niche specificity required.Thu 25 Jun“Learn AI” Is Bad Advice. Learn These InsteadVague advice to learn AI is dismissed in favor of six concrete skills that grow more valuable as AI improves. The episode argues that managing AI agents and local models is the most urgent technical skill, since most organizations will soon have dozens of half-working automations and no one capable of unifying them into a coherent system.Tue 23 JunGLM 5.2 Clearly Explained (and how to set it up)GLM 5.2, an open-source model from Zed AI, scores 81 points on Terminal Bench 2.1 and costs roughly 44 cents per run compared to $2.38 for Claude Opus 4.8 at equivalent token volumes, a five times price difference that compounds fast on large tasks. The episode walks through setting it up in Cursor by pasting the Zed AI API key into the OpenAI field and overriding the endpoint, with Open Router as a credit-based alternative.Mon 15 JunMaking $$$ with IOS appsGeorge quit his job at TJ Maxx to build iOS apps full-time using Rork, an AI coding tool requiring no prior programming knowledge, and turned that bet into nearly 200,000 dollars in total revenue from a single wrestling app.Sat 13 JunClaude Fable 5 is BANNED. What to do?When the US government sent Anthropic a letter on a Friday evening and Claude Fable 5 was immediately disabled for all users with no warning and no appeal, it exposed a structural vulnerability every cloud model shares: users rent access rather than own it.Thu 11 JunYou are using Claude Fable 5 wrongIn this episode, the host walks through a series of counterintuitive findings from early hands-on use of Claude 5, starting with the title claim that low effort mode outperforms high effort mode and should not be treated as a default. The episode covers a full AI video production pipeline that replaced work previously costing hundreds of thousands of dollars, using Claude 5 sub-agents to select takes and encode edits executed by FFmpeg.Tue 9 JunWhat are Agentic Loops?Ross Mike makes the case that agentic loops, where an agent checks and feeds its own output back into itself without human involvement, are widely misunderstood and misapplied. He points to AI figures Boris and Peter as prominent advocates of loops who operate without token budgets, with Peter burning 1.3 million dollars in tokens in a single month, making their approach impractical for users on 20 or 100 dollar monthly plans.Mon 8 JunBecome AI Native in less than 60 minsTheo Tabba defines an AI native organization as one where people manage agents, agents read and write to the company, and the company grows smarter over time, drawing a sharp contrast between casually using ChatGPT and being truly AI native.Sat 6 JunHermes Agent App Clearly Explained (and how to use it)In this episode, Alex Finn walks through the Hermes desktop app and explains why he considers it the best AI agent experience available, surpassing Claude and replacing earlier setups built around Telegram, Signal, and iMessage bots. He addresses the most common complaint, monthly costs often reported around one thousand dollars, tracing the problem to single-thread conversations that inflate token usage and arguing that the app's automatic session creation alone can cut costs three to four times.Thu 4 JunCodex Sites Clearly Explained (and how to use it)Codex Sites is an internal tool builder rather than a public deployment platform, meaning it ships without authentication, databases, or payment handling and cannot yet publish to a custom domain, though public domain deployment is expected soon.Tue 2 JunThe Next $100B Market: Selling To AI AgentsAI agents are rapidly displacing human beings as the primary customers and end users of software, and the speakers argue that almost no one is yet building for this machine-to-machine economy despite its potential to become a hundred-billion-dollar market. Current internet infrastructure assumes a human is doing the work, leaving agents without reliable identity, inboxes, memory, or spending capability, which means businesses need entirely rebuilt agent-native layers covering schemas, MCP tools, OAuth, and capability manifests.Tue 2 JunClearest Explanation of AI Agents as CustomersAI agents are rapidly replacing humans as the primary customers of the internet, and the infrastructure built for human browsing, searching, and buying is fundamentally mismatched for how agents discover, evaluate, and transact. Every step of the agent buying journey, from finding a service to paying for it and recommending it to other agents, lacks the identity systems, wallets, inboxes, and structured capability endpoints that agents require.Fri 22 MayGoogle's Biggest AI Announcements (I Was There)Logan Kilpatrick joined the show to break down Google's biggest announcements from I/O, describing Gemini 3.5 Flash as the most widely distributed day-one launch in Google's history, reaching 900 million users and repositioned as the primary model for the agent era with intelligence roughly at a Sonnet level.Fri 22 MayInside Google I/O with a DeepMind ExecLogan Kilpatrick of DeepMind joins to break down Google I/O, which was organized entirely around agents and what Jason calls a fundamental shift where useful work happens asynchronously without the user driving every step. Kilpatrick describes Gemini 2.5 Flash as the best model Google has ever shipped, now competing with Sonnet-level models on coding benchmarks and reaching 900 million users on launch day.