Uploaded May 2026 | Updated September 2026, 2 weeks ago
In this episode we sit down with 3 key members of the Replit Leadership Team.
Ghazi (Replit CRO) on how Replit is selling into the enterprise, the shift from seat-based to usage-based pricing, and what hiring looks like for non-traditional sales backgrounds.
Amjad (CEO) joins to talk about Jason Lemkin running SaaStr semi-autonomously with agents, why vices like laziness and being terminally online are becoming virtues, and why founders need to be athletes.
Michele (President & Head of AI) breaks down the philosophy behind the product roadmap, the agent harness, security as a first-class citizen, and why every Replit user gets the same agent regardless of plan. And Peter (AI engineer) - the surprise crowd favorite - talks model evaluation, the security scanner, and what "Vibench" means for the open source eval community.
Key themes: vibe coding going pro in 2026, why the org chart is becoming obsolete, how Replit dogfoods its own platform internally, and why distribution — not building — is now the hard part.
Timestamps
00:00 — Intro: Day 1 from SaaStr, why this year feels different from 2025
05:45 — Ghazi (CRO): enterprise demand, governed vibe coding, land-and-expand, building Bookie internally, the shift to usage-based pricing, and the upcoming London expansion
31:24 — Amjad (Founder & CEO): Jason Lemkin's semi-autonomous SaaStr, leading the market on agents, sunk cost fallacy, vices becoming virtues (lazy, terminally online, ADHD), flattening org charts, and platformizing agent creation
1:02:48 — Michele (President & Head of AI): the dream of empowering a billion creators, why "manager" is going away, security as non-negotiable, the agent harness as Replit's secret sauce, and why bringing your own keys isn't the answer
1:32:41 — Wrap-up reflections: stop over-engineering, start with one solvable task, and why UI still beats pure chat as an interface
1:42:55 — Surprise guest Peter (AI engineer): model evaluation, the security scanner, Vibench open source evals, and how the AI team picks models for different tasks
1:57:12 — Michele returns to drag Peter back to work, plus closing thoughts and what's next
In this episode we sit down with 3 key members of the Replit Leadership Team.
Ghazi (Replit CRO) on how Replit is selling into the enterprise, the shift from seat-based to usage-based pricing, and what hiring looks like for non-traditional sales backgrounds.
Amjad (CEO) joins to talk about Jason Lemkin running SaaStr semi-autonomously with agents, why vices like laziness and being terminally online are becoming virtues, and why founders need to be athletes.
Michele (President & Head of AI) breaks down the philosophy behind the product roadmap, the agent harness, security as a first-class citizen, and why every Replit user gets the same agent regardless of plan. And Peter (AI engineer) - the surprise crowd favorite - talks model evaluation, the security scanner, and what "Vibench" means for the open source eval community.
Key themes: vibe coding going pro in 2026, why the org chart is becoming obsolete, how Replit dogfoods its own platform internally, and why distribution — not building — is now the hard part.
Timestamps
00:00 — Intro: Day 1 from SaaStr, why this year feels different from 2025
05:45 — Ghazi (CRO): enterprise demand, governed vibe coding, land-and-expand, building Bookie internally, the shift to usage-based pricing, and the upcoming London expansion
31:24 — Amjad (Founder & CEO): Jason Lemkin's semi-autonomous SaaStr, leading the market on agents, sunk cost fallacy, vices becoming virtues (lazy, terminally online, ADHD), flattening org charts, and platformizing agent creation
1:02:48 — Michele (President & Head of AI): the dream of empowering a billion creators, why "manager" is going away, security as non-negotiable, the agent harness as Replit's secret sauce, and why bringing your own keys isn't the answer
1:32:41 — Wrap-up reflections: stop over-engineering, start with one solvable task, and why UI still beats pure chat as an interface
1:42:55 — Surprise guest Peter (AI engineer): model evaluation, the security scanner, Vibench open source evals, and how the AI team picks models for different tasks
1:57:12 — Michele returns to drag Peter back to work, plus closing thoughts and what's next
