StarZero All shows Try the podcast clipper →

Jan 23, 2026 · 42m · startup-ideas

I got a private lesson on Claude Cowork & Claude Code

Greg Isenberg · 10m spoken
0:00 / 0:00
▶ Watch on YouTube →

gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions

In this in-depth masterclass, Greg Isenberg sits down with Boris, the creator of Claude Code and co-creator of Claude Cowork, to demonstrate how Anthropic's autonomous AI agents automate desktop tasks, streamline workflows, and reshape productivity for technical and non-technical users alike.

How this conversation actually went

Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Greg holds 29.1% of the talking time here. How this is scored →

Greg as informed peer 2.8 Guest teaching 4.6 Guest disagreement 0.4 Greg pushing back 0.4
05100:0015:0030:001:00–3:26 · Greg as informed peer 3/10 The Origin and Unexpected Adoption of Claude Code Greg sets up the interview warmly and provides a solid platform analogy comparing early Claude Code adoption to unexpected App Store breakout apps. Boris agrees completely, reflecting on how early user behavior surprised Anthropic.3:27–8:29 · Greg as informed peer 2/10 Claude Cowork UI Overview and Local File Automation Boris gives an in-depth walkthrough of Claude Cowork, clarifying what agentic AI actually means beyond industry marketing hype and demonstrating reverse solicitation. Greg asks simple clarifying questions about file mounting.8:30–10:54 · Greg as informed peer 2/10 AI Safety Architecture and Spreadsheet Generation Boris details Anthropic's safety architecture, including mechanistic interpretability, sandboxed VMs, and prompt injection defenses during local file manipulation. Greg listens attentively and expresses amazement at the live spreadsheet generation.10:55–14:11 · Greg as informed peer 3/10 Browser Control and Parallel Workflow Execution Boris shows live browser automation in Google Sheets and shares the Anthropic internal pizza-ordering milestone. Greg responds by outlining a framework for auditing company-wide operational workflows for agentic automation.14:12–19:10 · Greg as informed peer 3/10 Multi-Clauding Workflows and Gmail Automation Greg raises a devil's advocate objection on behalf of YouTube skeptics regarding the sluggish speed of visual browser automation. Boris readily agrees and explains that running multiple Claudes in parallel completely offsets sequential latency.19:10–24:22 · Greg as informed peer 3/10 Non-Technical Adoption, Skills, and Tool Customization Boris explains the technical mechanics of Skills and contrasts how software engineers demand extreme customizability while everyday Cowork users benefit from out-of-the-box simplicity. Greg correctly deduces that front-loading custom skill creation improves agent outcomes.24:22–28:41 · Greg as informed peer 4/10 Future Predictions and the Exponential Trajectory of AI When asked for a one-year forecast, Boris rejects multi-year linear planning because human brains struggle with exponential curves, noting that Claude already writes 100% of his code. Greg responds with his own thesis on vertical agent hiring and specialized digital teammates.28:42–40:06 · Greg as informed peer 2/10 Boris's Viral Masterclass on Claude Code Best Practices Greg shares Boris's viral Claude Code thread on screen while Boris delivers practical masterclass tips on Opus 4.5 efficiency, CLAUDE.md compounding workflows, and verification loops using browser tooling.1:00–3:26 · Guest teaching 2/10 The Origin and Unexpected Adoption of Claude Code Greg sets up the interview warmly and provides a solid platform analogy comparing early Claude Code adoption to unexpected App Store breakout apps. Boris agrees completely, reflecting on how early user behavior surprised Anthropic.3:27–8:29 · Guest teaching 5/10 Claude Cowork UI Overview and Local File Automation Boris gives an in-depth walkthrough of Claude Cowork, clarifying what agentic AI actually means beyond industry marketing hype and demonstrating reverse solicitation. Greg asks simple clarifying questions about file mounting.8:30–10:54 · Guest teaching 6/10 AI Safety Architecture and Spreadsheet Generation Boris details Anthropic's safety architecture, including mechanistic interpretability, sandboxed VMs, and prompt injection defenses during local file manipulation. Greg listens attentively and expresses amazement at the live spreadsheet generation.10:55–14:11 · Guest teaching 4/10 Browser Control and Parallel Workflow Execution Boris shows live browser automation in Google Sheets and shares the Anthropic internal pizza-ordering milestone. Greg responds by outlining a framework for auditing company-wide operational workflows for agentic automation.14:12–19:10 · Guest teaching 4/10 Multi-Clauding Workflows and Gmail Automation Greg raises a devil's advocate objection on behalf of YouTube skeptics regarding the sluggish speed of visual browser automation. Boris readily agrees and explains that running multiple Claudes in parallel completely offsets sequential latency.19:10–24:22 · Guest teaching 5/10 Non-Technical Adoption, Skills, and Tool Customization Boris explains the technical mechanics of Skills and contrasts how software engineers demand extreme customizability while everyday Cowork users benefit from out-of-the-box simplicity. Greg correctly deduces that front-loading custom skill creation improves agent outcomes.24:22–28:41 · Guest teaching 5/10 Future Predictions and the Exponential Trajectory of AI When asked for a one-year forecast, Boris rejects multi-year linear planning because human brains struggle with exponential curves, noting that Claude already writes 100% of his code. Greg responds with his own thesis on vertical agent hiring and specialized digital teammates.28:42–40:06 · Guest teaching 6/10 Boris's Viral Masterclass on Claude Code Best Practices Greg shares Boris's viral Claude Code thread on screen while Boris delivers practical masterclass tips on Opus 4.5 efficiency, CLAUDE.md compounding workflows, and verification loops using browser tooling.1:00–3:26 · Guest disagreement 0/10 The Origin and Unexpected Adoption of Claude Code Greg sets up the interview warmly and provides a solid platform analogy comparing early Claude Code adoption to unexpected App Store breakout apps. Boris agrees completely, reflecting on how early user behavior surprised Anthropic.3:27–8:29 · Guest disagreement 1/10 Claude Cowork UI Overview and Local File Automation Boris gives an in-depth walkthrough of Claude Cowork, clarifying what agentic AI actually means beyond industry marketing hype and demonstrating reverse solicitation. Greg asks simple clarifying questions about file mounting.8:30–10:54 · Guest disagreement 0/10 AI Safety Architecture and Spreadsheet Generation Boris details Anthropic's safety architecture, including mechanistic interpretability, sandboxed VMs, and prompt injection defenses during local file manipulation. Greg listens attentively and expresses amazement at the live spreadsheet generation.10:55–14:11 · Guest disagreement 0/10 Browser Control and Parallel Workflow Execution Boris shows live browser automation in Google Sheets and shares the Anthropic internal pizza-ordering milestone. Greg responds by outlining a framework for auditing company-wide operational workflows for agentic automation.14:12–19:10 · Guest disagreement 1/10 Multi-Clauding Workflows and Gmail Automation Greg raises a devil's advocate objection on behalf of YouTube skeptics regarding the sluggish speed of visual browser automation. Boris readily agrees and explains that running multiple Claudes in parallel completely offsets sequential latency.19:10–24:22 · Guest disagreement 0/10 Non-Technical Adoption, Skills, and Tool Customization Boris explains the technical mechanics of Skills and contrasts how software engineers demand extreme customizability while everyday Cowork users benefit from out-of-the-box simplicity. Greg correctly deduces that front-loading custom skill creation improves agent outcomes.24:22–28:41 · Guest disagreement 1/10 Future Predictions and the Exponential Trajectory of AI When asked for a one-year forecast, Boris rejects multi-year linear planning because human brains struggle with exponential curves, noting that Claude already writes 100% of his code. Greg responds with his own thesis on vertical agent hiring and specialized digital teammates.28:42–40:06 · Guest disagreement 0/10 Boris's Viral Masterclass on Claude Code Best Practices Greg shares Boris's viral Claude Code thread on screen while Boris delivers practical masterclass tips on Opus 4.5 efficiency, CLAUDE.md compounding workflows, and verification loops using browser tooling.1:00–3:26 · Greg pushing back 0/10 The Origin and Unexpected Adoption of Claude Code Greg sets up the interview warmly and provides a solid platform analogy comparing early Claude Code adoption to unexpected App Store breakout apps. Boris agrees completely, reflecting on how early user behavior surprised Anthropic.3:27–8:29 · Greg pushing back 0/10 Claude Cowork UI Overview and Local File Automation Boris gives an in-depth walkthrough of Claude Cowork, clarifying what agentic AI actually means beyond industry marketing hype and demonstrating reverse solicitation. Greg asks simple clarifying questions about file mounting.8:30–10:54 · Greg pushing back 0/10 AI Safety Architecture and Spreadsheet Generation Boris details Anthropic's safety architecture, including mechanistic interpretability, sandboxed VMs, and prompt injection defenses during local file manipulation. Greg listens attentively and expresses amazement at the live spreadsheet generation.10:55–14:11 · Greg pushing back 0/10 Browser Control and Parallel Workflow Execution Boris shows live browser automation in Google Sheets and shares the Anthropic internal pizza-ordering milestone. Greg responds by outlining a framework for auditing company-wide operational workflows for agentic automation.14:12–19:10 · Greg pushing back 2/10 Multi-Clauding Workflows and Gmail Automation Greg raises a devil's advocate objection on behalf of YouTube skeptics regarding the sluggish speed of visual browser automation. Boris readily agrees and explains that running multiple Claudes in parallel completely offsets sequential latency.19:10–24:22 · Greg pushing back 0/10 Non-Technical Adoption, Skills, and Tool Customization Boris explains the technical mechanics of Skills and contrasts how software engineers demand extreme customizability while everyday Cowork users benefit from out-of-the-box simplicity. Greg correctly deduces that front-loading custom skill creation improves agent outcomes.24:22–28:41 · Greg pushing back 1/10 Future Predictions and the Exponential Trajectory of AI When asked for a one-year forecast, Boris rejects multi-year linear planning because human brains struggle with exponential curves, noting that Claude already writes 100% of his code. Greg responds with his own thesis on vertical agent hiring and specialized digital teammates.28:42–40:06 · Greg pushing back 0/10 Boris's Viral Masterclass on Claude Code Best Practices Greg shares Boris's viral Claude Code thread on screen while Boris delivers practical masterclass tips on Opus 4.5 efficiency, CLAUDE.md compounding workflows, and verification loops using browser tooling.

speaking balance: gold is Greg, purple is the guest (3 minute bins)

0:00 · Greg 61.9% · guest 38.1%0:00 · Greg 61.9% · guest 38.1%3:00 · Greg 15% · guest 85%3:00 · Greg 15% · guest 85%6:00 · Greg 11.7% · guest 88.3%6:00 · Greg 11.7% · guest 88.3%9:00 · Greg 14.2% · guest 85.8%9:00 · Greg 14.2% · guest 85.8%12:00 · Greg 34.2% · guest 65.8%12:00 · Greg 34.2% · guest 65.8%15:00 · Greg 24.6% · guest 75.4%15:00 · Greg 24.6% · guest 75.4%18:00 · Greg 16.1% · guest 83.9%18:00 · Greg 16.1% · guest 83.9%21:00 · Greg 23.3% · guest 76.7%21:00 · Greg 23.3% · guest 76.7%24:00 · Greg 17.1% · guest 82.9%24:00 · Greg 17.1% · guest 82.9%27:00 · Greg 64.6% · guest 35.4%27:00 · Greg 64.6% · guest 35.4%30:00 · Greg 20.2% · guest 79.8%30:00 · Greg 20.2% · guest 79.8%33:00 · Greg 44.7% · guest 55.3%33:00 · Greg 44.7% · guest 55.3%36:00 · Greg 20.9% · guest 79.1%36:00 · Greg 20.9% · guest 79.1%39:00 · Greg 37.6% · guest 62.4%39:00 · Greg 37.6% · guest 62.4%42:00 · Greg 84.9% · guest 15.1%42:00 · Greg 84.9% · guest 15.1%
Sharpest disagreement ▶ 24:48 Boris rejects linear 12-month prediction framing

Boris dismisses long-range forecasting questions by arguing that human brains cannot intuitively grasp exponential AI curves, explaining he only plans on a one-week horizon.

 Hardest push from Greg ▶ 16:57 Greg channels audience skepticism over UI execution speed

Greg plays devil's advocate against the live demo, pushing Boris to defend whether the tool is too slow compared to manual human execution.

 Biggest teaching moment ▶ 34:00 Boris explains why Opus 4.5 is actually cheaper

Boris educates listeners on the counterintuitive economics of LLMs, demonstrating that a smarter model requires fewer steering tokens and finishes tasks at a lower total cost than smaller models.

 Greg holds their own ▶ 26:54 Greg outlines the vertical digital teammate thesis

Greg demonstrates market insight by building on Boris's points and proposing that Cowork will evolve into verticalized digital role staffing, analogous to specialized job boards.
the scores for every segment, with the reasoning behind each
ChapterTopicGreg as informed peerGuest teachingGuest disagreementGreg pushing backWhy
The Origin and Unexpected Adoption of Claude Code 3200 Greg sets up the interview warmly and provides a solid platform analogy comparing early Claude Code adoption to unexpected App Store breakout apps. Boris agrees completely, reflecting on how early user behavior surprised Anthropic.
Claude Cowork UI Overview and Local File Automation 2510 Boris gives an in-depth walkthrough of Claude Cowork, clarifying what agentic AI actually means beyond industry marketing hype and demonstrating reverse solicitation. Greg asks simple clarifying questions about file mounting.
AI Safety Architecture and Spreadsheet Generation 2600 Boris details Anthropic's safety architecture, including mechanistic interpretability, sandboxed VMs, and prompt injection defenses during local file manipulation. Greg listens attentively and expresses amazement at the live spreadsheet generation.
Browser Control and Parallel Workflow Execution 3400 Boris shows live browser automation in Google Sheets and shares the Anthropic internal pizza-ordering milestone. Greg responds by outlining a framework for auditing company-wide operational workflows for agentic automation.
Multi-Clauding Workflows and Gmail Automation 3412 Greg raises a devil's advocate objection on behalf of YouTube skeptics regarding the sluggish speed of visual browser automation. Boris readily agrees and explains that running multiple Claudes in parallel completely offsets sequential latency.
Non-Technical Adoption, Skills, and Tool Customization 3500 Boris explains the technical mechanics of Skills and contrasts how software engineers demand extreme customizability while everyday Cowork users benefit from out-of-the-box simplicity. Greg correctly deduces that front-loading custom skill creation improves agent outcomes.
Future Predictions and the Exponential Trajectory of AI 4511 When asked for a one-year forecast, Boris rejects multi-year linear planning because human brains struggle with exponential curves, noting that Claude already writes 100% of his code. Greg responds with his own thesis on vertical agent hiring and specialized digital teammates.
Boris's Viral Masterclass on Claude Code Best Practices 2600 Greg shares Boris's viral Claude Code thread on screen while Boris delivers practical masterclass tips on Opus 4.5 efficiency, CLAUDE.md compounding workflows, and verification loops using browser tooling.

Statements from this episode (0)

Nothing in this episode matches those filters. clear them

Made with StarZero

Turn any episode into a week of clips.

This entire site, over 300 episodes transcribed, diarized, checked and made playable, runs on the StarZero media pipeline. Drop in your own episode and the podcast clipper finds the moments worth sharing, cuts them, captions them, and reframes them for every feed.

Try the podcast clipper Learn more