Temporal CEO on AI Agents & The Future of Software | Deep Dives with a16z

Video summary

Temporal CEO Samar Abbas says durable execution will underpin agent swarms, with OpenAI Codex and 150,000 actions per second already in focus.

Temporal CEO Samar Abbas explains how durable execution preserves workflow state and resumes code after failures, drawing on Uber’s Cadence origins in a microservices system with more services than engineers. Uber used it for three-day bank retries and loyalty points; event histories let the team replay corrupted workflows after a bug erased customer balances. Abbas says Temporal Cloud offers 99.99% operational availability and can handle sudden spikes of 150,000 actions per second. As agents shift from short interactive sessions to long-running background work, Temporal supports OpenAI Codex executions at scale, parallel deep research, and recoverable human-in-the-loop workflows. He predicts specialized agents will need durable RPC, with Temporal’s Project Nexus aiming to help standardize asynchronous tool calls. The discussion also covers real-time context gathering, enterprise guardrails, why SaaS is not dead, and the CEO transition from cofounder Max to Abbas as the company prioritized execution. Abbas’s advice to founders: focus on customer value, build resilient margins, and invest for the opportunity ahead.

Chapters

1. 0:00Introduction: Temporal resumes failed executions and processes restaurant orders exactly once
2. 4:03Temporal’s Origin Story: Uber’s Cadence, three-day retries, and loyalty-point recovery
3. 11:14Why Agents Raise the Stakes: AI-generated apps make reliability and operations more important
4. 16:00Specialized Agents Need Durable RPC: From Claude Code’s sandbox to travel agents and hundreds of agents
5. 19:54Specialized Agents Need Durable RPC: OpenAI Codex and engineers delegating 15 background tasks
6. 25:20Deep Research Agents: Durable state, parallel research, and 150,000 actions per second
7. 30:58Execution Histories as a Superpower: AI agents dwarf Snap’s scale, while workflow histories enable auditability and analytics
8. 35:03Execution Histories as a Superpower: APIs retain value as agents multiply, and Temporal integrates with OpenAI’s Agentic SDK
9. 39:04Minimal Viable Long-Running Agent Architecture: Sandboxes, prompt evaluation, and observability alongside enterprise systems of record
10. 45:07Context Engineering at Scale: Real-time retrieval from APIs, Slack, and Google Docs drives large-payload and durable-RPC needs
11. 52:40Where Value Accrues: Jensen’s five-layer cake, Abridge and Harvey, and Samar’s CEO transition at Temporal
12. 57:26Where Value Accrues: Temporal’s post-2021 margin recovery, customer focus, and R&D investment

What the full analysis includes

12 chapters · 76 key points · 12 section summaries · mind map

The remaining 73 key points, section summaries and mind map are in the full analysis. Analyze this video to see everything →

This video is 1.1 hours. The Free plan analyzes videos up to 30 min; plans from Plus cover videos up to 2 hours.