Dev Interrupted Moltbook, Rent-a-Human, Super Agents & Connectivity Benchmark Report | ft. Gary Lerhaupt
12 snips
Feb 7, 2026 Gary Lerhaupt, VP of Product Architecture at Salesforce who focuses on agentic systems and AI orchestration, joins to discuss AI's wild new behaviors. Conversation covers Moltbook's agent societies and Rent-a-Human's agent-to-human marketplaces. They unpack risks like OpenClaw, the shift to specialist agents, and Salesforce's Connectivity Benchmark and AgentForce approaches to orchestration and governance.
AI Snips
Chapters
Transcript
Episode notes
Agent Cascades Increase Attack Surface
- Running agent cascades like OpenClaw on personal devices creates acute security and prompt-injection risks.
- The rapid emergence of autonomous agent tooling demands new guardrails and threat models.
Agentic Operating Model Is Here
- Gartner predicts a rapid rise of agentic features in enterprise apps, indicating the agentic operating model is arriving.
- LLMs excel at iterative loops (observe, orient, decide, act) that redefine knowledge work.
Open Source Must Serve Agents
- Vibe coding and agent use threaten traditional open source discovery and maintenance dynamics.
- Projects must optimize for agent experience, not just human docs and community size.