StarZero All shows Try the podcast clipper →

Mar 26, 2026 · 46m · startup-ideas

Paperclip: Hire AI Agents Like Employees (Live Demo)

Greg Isenberg · 5m 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

Greg Isenberg interviews Dota, creator of the open-source orchestrator Paperclip, to explore how founders can construct, manage, and scale zero-human companies using hierarchical AI agent teams. Through live demonstrations, Dota reveals how to configure agent memory, assign modular skills, enforce quality assurance loops, and automate organizational workflows.

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 13.2% of the talking time here. How this is scored →

Greg as informed peer 3.9 Guest teaching 4.3 Guest disagreement 0.2 Greg pushing back 0.7
05100:0015:0030:0045:000:53–4:27 · Greg as informed peer 3/10 Episode Objectives and Dota's Avatar Persona Greg sets the stage and playfully inquires about Dota's avatar persona and anonymity. Dota explains the core philosophy behind Paperclip as an orchestration layer for zero-human companies while citing Naval Ravikant.4:28–7:50 · Greg as informed peer 3/10 Company Inception and CEO Agent Setup Greg participates by picking a business idea ('Moolah') from his idea browser. Dota takes over the walkthrough, demonstrating how to initialize a company, configure frontier models for the CEO agent, and leverage free OpenRouter models.7:51–11:12 · Greg as informed peer 4/10 Initial Delegation, Budget Tracking, and Hiring Approval Dota explains Paperclip's origin story of token budget blowouts with multi-window Claude Code sessions and details the issue-tracking model. Greg synthesizes the discussion by identifying the high-level mindset shift from 'jobs to be done' to 'hiring agents.'11:12–17:14 · Greg as informed peer 2/10 Agent Persona Configuration, Memory, and the Memento Analogy Dota educates the audience on agent memory systems (PARA) and delivers the 'Memento' analogy for agent heartbeats and prompt iterations. Greg actively listens and prompts for deeper configuration details.17:15–21:02 · Greg as informed peer 5/10 Extending Agents with Skills and Security Considerations Greg raises critical questions regarding the security risks of installing third-party agent skills and notes that GitHub star counts do not guarantee safety. Dota concedes the point and explains the limits of current audit badges on skills.sh.21:03–25:58 · Greg as informed peer 4/10 Contextual Artifacts, Brand Guidelines, and Video Generation Greg asks practical questions about ensuring quality video output from Remotion skills and playfully probes whether agents can be fired. Dota demonstrates artifact linking, brand guidelines, and the terminate button.25:58–30:11 · Greg as informed peer 6/10 Harness Engineering, Feedback Loops, and Inscribing Taste Dota breaks down harness engineering and the necessity of instilling human taste into agent prompts. Greg draws a parallel to classical founder leadership, noting that communicating taste and values remains unchanged whether managing humans or agents.30:12–36:41 · Greg as informed peer 3/10 Building Internal Org Charts and Automated Routines Dota walks through his own Paperclip company org chart, evals engineering, and automated recurring routines by configuring a live GitHub-to-Discord update agent. Greg observes with high interest.36:41–38:57 · Greg as informed peer 3/10 Real-World Use Cases and Diverse Applications Greg asks if Paperclip itself is the most successful project built on the platform. Dota reframes success around revenue versus GitHub stars, citing real-world examples including roofers, dentists, and security auditors.38:58–43:26 · Greg as informed peer 5/10 Importable Company Templates and Upcoming Maximizer Mode Dota previews importable company templates and the autonomous 'Maximizer Mode'. Greg astutely frames this capability as the agentic equivalent of 'acquihiring' proven teams into an existing company.43:26–46:40 · Greg as informed peer 5/10 The Team Behind Paperclip, Virality, and Future Outlook Greg identifies co-founder Devin Foley and validates the team's pedigree based on their previous startup relationship. Dota discusses the 'bitter lesson' in AI and why orchestration layers will retain staying power.0:53–4:27 · Guest teaching 4/10 Episode Objectives and Dota's Avatar Persona Greg sets the stage and playfully inquires about Dota's avatar persona and anonymity. Dota explains the core philosophy behind Paperclip as an orchestration layer for zero-human companies while citing Naval Ravikant.4:28–7:50 · Guest teaching 5/10 Company Inception and CEO Agent Setup Greg participates by picking a business idea ('Moolah') from his idea browser. Dota takes over the walkthrough, demonstrating how to initialize a company, configure frontier models for the CEO agent, and leverage free OpenRouter models.7:51–11:12 · Guest teaching 4/10 Initial Delegation, Budget Tracking, and Hiring Approval Dota explains Paperclip's origin story of token budget blowouts with multi-window Claude Code sessions and details the issue-tracking model. Greg synthesizes the discussion by identifying the high-level mindset shift from 'jobs to be done' to 'hiring agents.'11:12–17:14 · Guest teaching 6/10 Agent Persona Configuration, Memory, and the Memento Analogy Dota educates the audience on agent memory systems (PARA) and delivers the 'Memento' analogy for agent heartbeats and prompt iterations. Greg actively listens and prompts for deeper configuration details.17:15–21:02 · Guest teaching 3/10 Extending Agents with Skills and Security Considerations Greg raises critical questions regarding the security risks of installing third-party agent skills and notes that GitHub star counts do not guarantee safety. Dota concedes the point and explains the limits of current audit badges on skills.sh.21:03–25:58 · Guest teaching 4/10 Contextual Artifacts, Brand Guidelines, and Video Generation Greg asks practical questions about ensuring quality video output from Remotion skills and playfully probes whether agents can be fired. Dota demonstrates artifact linking, brand guidelines, and the terminate button.25:58–30:11 · Guest teaching 4/10 Harness Engineering, Feedback Loops, and Inscribing Taste Dota breaks down harness engineering and the necessity of instilling human taste into agent prompts. Greg draws a parallel to classical founder leadership, noting that communicating taste and values remains unchanged whether managing humans or agents.30:12–36:41 · Guest teaching 5/10 Building Internal Org Charts and Automated Routines Dota walks through his own Paperclip company org chart, evals engineering, and automated recurring routines by configuring a live GitHub-to-Discord update agent. Greg observes with high interest.36:41–38:57 · Guest teaching 4/10 Real-World Use Cases and Diverse Applications Greg asks if Paperclip itself is the most successful project built on the platform. Dota reframes success around revenue versus GitHub stars, citing real-world examples including roofers, dentists, and security auditors.38:58–43:26 · Guest teaching 4/10 Importable Company Templates and Upcoming Maximizer Mode Dota previews importable company templates and the autonomous 'Maximizer Mode'. Greg astutely frames this capability as the agentic equivalent of 'acquihiring' proven teams into an existing company.43:26–46:40 · Guest teaching 4/10 The Team Behind Paperclip, Virality, and Future Outlook Greg identifies co-founder Devin Foley and validates the team's pedigree based on their previous startup relationship. Dota discusses the 'bitter lesson' in AI and why orchestration layers will retain staying power.0:53–4:27 · Guest disagreement 0/10 Episode Objectives and Dota's Avatar Persona Greg sets the stage and playfully inquires about Dota's avatar persona and anonymity. Dota explains the core philosophy behind Paperclip as an orchestration layer for zero-human companies while citing Naval Ravikant.4:28–7:50 · Guest disagreement 0/10 Company Inception and CEO Agent Setup Greg participates by picking a business idea ('Moolah') from his idea browser. Dota takes over the walkthrough, demonstrating how to initialize a company, configure frontier models for the CEO agent, and leverage free OpenRouter models.7:51–11:12 · Guest disagreement 0/10 Initial Delegation, Budget Tracking, and Hiring Approval Dota explains Paperclip's origin story of token budget blowouts with multi-window Claude Code sessions and details the issue-tracking model. Greg synthesizes the discussion by identifying the high-level mindset shift from 'jobs to be done' to 'hiring agents.'11:12–17:14 · Guest disagreement 0/10 Agent Persona Configuration, Memory, and the Memento Analogy Dota educates the audience on agent memory systems (PARA) and delivers the 'Memento' analogy for agent heartbeats and prompt iterations. Greg actively listens and prompts for deeper configuration details.17:15–21:02 · Guest disagreement 1/10 Extending Agents with Skills and Security Considerations Greg raises critical questions regarding the security risks of installing third-party agent skills and notes that GitHub star counts do not guarantee safety. Dota concedes the point and explains the limits of current audit badges on skills.sh.21:03–25:58 · Guest disagreement 0/10 Contextual Artifacts, Brand Guidelines, and Video Generation Greg asks practical questions about ensuring quality video output from Remotion skills and playfully probes whether agents can be fired. Dota demonstrates artifact linking, brand guidelines, and the terminate button.25:58–30:11 · Guest disagreement 0/10 Harness Engineering, Feedback Loops, and Inscribing Taste Dota breaks down harness engineering and the necessity of instilling human taste into agent prompts. Greg draws a parallel to classical founder leadership, noting that communicating taste and values remains unchanged whether managing humans or agents.30:12–36:41 · Guest disagreement 0/10 Building Internal Org Charts and Automated Routines Dota walks through his own Paperclip company org chart, evals engineering, and automated recurring routines by configuring a live GitHub-to-Discord update agent. Greg observes with high interest.36:41–38:57 · Guest disagreement 1/10 Real-World Use Cases and Diverse Applications Greg asks if Paperclip itself is the most successful project built on the platform. Dota reframes success around revenue versus GitHub stars, citing real-world examples including roofers, dentists, and security auditors.38:58–43:26 · Guest disagreement 0/10 Importable Company Templates and Upcoming Maximizer Mode Dota previews importable company templates and the autonomous 'Maximizer Mode'. Greg astutely frames this capability as the agentic equivalent of 'acquihiring' proven teams into an existing company.43:26–46:40 · Guest disagreement 0/10 The Team Behind Paperclip, Virality, and Future Outlook Greg identifies co-founder Devin Foley and validates the team's pedigree based on their previous startup relationship. Dota discusses the 'bitter lesson' in AI and why orchestration layers will retain staying power.0:53–4:27 · Greg pushing back 1/10 Episode Objectives and Dota's Avatar Persona Greg sets the stage and playfully inquires about Dota's avatar persona and anonymity. Dota explains the core philosophy behind Paperclip as an orchestration layer for zero-human companies while citing Naval Ravikant.4:28–7:50 · Greg pushing back 0/10 Company Inception and CEO Agent Setup Greg participates by picking a business idea ('Moolah') from his idea browser. Dota takes over the walkthrough, demonstrating how to initialize a company, configure frontier models for the CEO agent, and leverage free OpenRouter models.7:51–11:12 · Greg pushing back 0/10 Initial Delegation, Budget Tracking, and Hiring Approval Dota explains Paperclip's origin story of token budget blowouts with multi-window Claude Code sessions and details the issue-tracking model. Greg synthesizes the discussion by identifying the high-level mindset shift from 'jobs to be done' to 'hiring agents.'11:12–17:14 · Greg pushing back 0/10 Agent Persona Configuration, Memory, and the Memento Analogy Dota educates the audience on agent memory systems (PARA) and delivers the 'Memento' analogy for agent heartbeats and prompt iterations. Greg actively listens and prompts for deeper configuration details.17:15–21:02 · Greg pushing back 3/10 Extending Agents with Skills and Security Considerations Greg raises critical questions regarding the security risks of installing third-party agent skills and notes that GitHub star counts do not guarantee safety. Dota concedes the point and explains the limits of current audit badges on skills.sh.21:03–25:58 · Greg pushing back 2/10 Contextual Artifacts, Brand Guidelines, and Video Generation Greg asks practical questions about ensuring quality video output from Remotion skills and playfully probes whether agents can be fired. Dota demonstrates artifact linking, brand guidelines, and the terminate button.25:58–30:11 · Greg pushing back 1/10 Harness Engineering, Feedback Loops, and Inscribing Taste Dota breaks down harness engineering and the necessity of instilling human taste into agent prompts. Greg draws a parallel to classical founder leadership, noting that communicating taste and values remains unchanged whether managing humans or agents.30:12–36:41 · Greg pushing back 0/10 Building Internal Org Charts and Automated Routines Dota walks through his own Paperclip company org chart, evals engineering, and automated recurring routines by configuring a live GitHub-to-Discord update agent. Greg observes with high interest.36:41–38:57 · Greg pushing back 1/10 Real-World Use Cases and Diverse Applications Greg asks if Paperclip itself is the most successful project built on the platform. Dota reframes success around revenue versus GitHub stars, citing real-world examples including roofers, dentists, and security auditors.38:58–43:26 · Greg pushing back 0/10 Importable Company Templates and Upcoming Maximizer Mode Dota previews importable company templates and the autonomous 'Maximizer Mode'. Greg astutely frames this capability as the agentic equivalent of 'acquihiring' proven teams into an existing company.43:26–46:40 · Greg pushing back 0/10 The Team Behind Paperclip, Virality, and Future Outlook Greg identifies co-founder Devin Foley and validates the team's pedigree based on their previous startup relationship. Dota discusses the 'bitter lesson' in AI and why orchestration layers will retain staying power.

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

0:00 · Greg 52.2% · guest 47.8%0:00 · Greg 52.2% · guest 47.8%3:00 · Greg 15.2% · guest 84.8%3:00 · Greg 15.2% · guest 84.8%6:00 · Greg 3.5% · guest 96.5%6:00 · Greg 3.5% · guest 96.5%9:00 · Greg 5.6% · guest 94.4%9:00 · Greg 5.6% · guest 94.4%12:00 · Greg 5.9% · guest 94.1%12:00 · Greg 5.9% · guest 94.1%15:00 · Greg 0% · guest 100%15:00 · Greg 0% · guest 100%18:00 · Greg 18.3% · guest 81.7%18:00 · Greg 18.3% · guest 81.7%21:00 · Greg 17.3% · guest 82.7%21:00 · Greg 17.3% · guest 82.7%24:00 · Greg 8.4% · guest 91.6%24:00 · Greg 8.4% · guest 91.6%27:00 · Greg 9.4% · guest 90.6%27:00 · Greg 9.4% · guest 90.6%30:00 · Greg 19.1% · guest 80.9%30:00 · Greg 19.1% · guest 80.9%33:00 · Greg 0.6% · guest 99.4%33:00 · Greg 0.6% · guest 99.4%36:00 · Greg 4.2% · guest 95.8%36:00 · Greg 4.2% · guest 95.8%39:00 · Greg 2% · guest 98%39:00 · Greg 2% · guest 98%42:00 · Greg 25.7% · guest 74.3%42:00 · Greg 25.7% · guest 74.3%45:00 · Greg 34.1% · guest 65.9%45:00 · Greg 34.1% · guest 65.9%
Sharpest disagreement ▶ 38:38 Dota rejects star metrics to define company success

Dota counters Greg's premise that Paperclip is the most successful project by highlighting that Paperclip lacks revenue, making paying businesses more genuinely successful.

 Hardest push from Greg ▶ 18:49 Greg questions the security and malicious risks of skills

Greg directly challenges the assumption that popular open-source skills on skills.sh are safe, pushing Dota to address how users avoid malicious code injection.

 Biggest teaching moment ▶ 14:09 Dota introduces the Memento framework for agent heartbeats

Dota provides a memorable mental model comparing stateless AI agents waking up on each heartbeat to the amnesiac protagonist of Memento requiring persistent instruction notes.

 Greg holds their own ▶ 29:42 Greg synthesizes agent taste configuration with executive leadership

Greg demonstrates strong domain intuition by re-anchoring Dota's discussion of harness engineering and taste into the classic definition of great managerial and founder leadership.
the scores for every segment, with the reasoning behind each
ChapterTopicGreg as informed peerGuest teachingGuest disagreementGreg pushing backWhy
Episode Objectives and Dota's Avatar Persona 3401 Greg sets the stage and playfully inquires about Dota's avatar persona and anonymity. Dota explains the core philosophy behind Paperclip as an orchestration layer for zero-human companies while citing Naval Ravikant.
Company Inception and CEO Agent Setup 3500 Greg participates by picking a business idea ('Moolah') from his idea browser. Dota takes over the walkthrough, demonstrating how to initialize a company, configure frontier models for the CEO agent, and leverage free OpenRouter models.
Initial Delegation, Budget Tracking, and Hiring Approval 4400 Dota explains Paperclip's origin story of token budget blowouts with multi-window Claude Code sessions and details the issue-tracking model. Greg synthesizes the discussion by identifying the high-level mindset shift from 'jobs to be done' to 'hiring agents.'
Agent Persona Configuration, Memory, and the Memento Analogy 2600 Dota educates the audience on agent memory systems (PARA) and delivers the 'Memento' analogy for agent heartbeats and prompt iterations. Greg actively listens and prompts for deeper configuration details.
Extending Agents with Skills and Security Considerations 5313 Greg raises critical questions regarding the security risks of installing third-party agent skills and notes that GitHub star counts do not guarantee safety. Dota concedes the point and explains the limits of current audit badges on skills.sh.
Contextual Artifacts, Brand Guidelines, and Video Generation 4402 Greg asks practical questions about ensuring quality video output from Remotion skills and playfully probes whether agents can be fired. Dota demonstrates artifact linking, brand guidelines, and the terminate button.
Harness Engineering, Feedback Loops, and Inscribing Taste 6401 Dota breaks down harness engineering and the necessity of instilling human taste into agent prompts. Greg draws a parallel to classical founder leadership, noting that communicating taste and values remains unchanged whether managing humans or agents.
Building Internal Org Charts and Automated Routines 3500 Dota walks through his own Paperclip company org chart, evals engineering, and automated recurring routines by configuring a live GitHub-to-Discord update agent. Greg observes with high interest.
Real-World Use Cases and Diverse Applications 3411 Greg asks if Paperclip itself is the most successful project built on the platform. Dota reframes success around revenue versus GitHub stars, citing real-world examples including roofers, dentists, and security auditors.
Importable Company Templates and Upcoming Maximizer Mode 5400 Dota previews importable company templates and the autonomous 'Maximizer Mode'. Greg astutely frames this capability as the agentic equivalent of 'acquihiring' proven teams into an existing company.
The Team Behind Paperclip, Virality, and Future Outlook 5400 Greg identifies co-founder Devin Foley and validates the team's pedigree based on their previous startup relationship. Dota discusses the 'bitter lesson' in AI and why orchestration layers will retain staying power.

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