AI + a16z Durable Execution and the Infrastructure Powering AI Agents
67 snips
Feb 19, 2026 Samar Abbas, CEO of Temporal and creator of the durable-execution system that began as Cadence at Uber. He explains why long-running AI agents need persistent state and recoverability. He describes how Temporal powers production workloads like OpenAI Codex and Snap, and why background, multi-agent systems create new distributed systems challenges at internet scale.
AI Snips
Chapters
Transcript
Episode notes
Replace Long Kafka Retries With Workflows
- Use durable workflows instead of retrying messages endlessly in streaming systems like Kafka.
- Start a workflow with a multi-day retry policy to manage long-running external call reliability.
Agentic Apps Create A New Orchestration Need
- Agentic apps turn models into planners that invoke tools and run asynchronously, creating many more developer-driven apps.
- Temporal's durable execution matches that shift by handling long-lived state, recoverability, and scale for these apps.
From Interactive To Long-Running Agents
- Agents are shifting from short interactive sessions to long-running background workers that need durability.
- Temporal maps naturally to agent loops and supports multi-agent swarms running concurrently.