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

a16z Deep Dives · 2026-02-19 · 63 分鐘
https://www.youtube.com/watch?v=NaIiiON5Sj4影片總結
Temporal CEO Samar Abbas says durable execution will underpin long-running AI agents, with cloud handling 150,000 actions per second.
Temporal CEO Samar Abbas explains how durable execution resumes code after failures with its state intact, a capability born from Uber’s microservices sprawl and its Cadence predecessor. He recounts how Uber used workflows for three-day bank retries and loyalty points—and replayed corrupted workflows after a bug erased customer credits. Temporal Cloud offers 99.999% operational SLA, regional failover, and capacity for 150,000 actions per second. Abbas argues that AI is shifting from short interactive sessions to background coding and research agents that need recoverable state, durable RPC, and coordination across specialized agents. Execution histories also provide audit trails and observability, while real-time context increasingly comes from APIs, Slack, and Google Docs. He sees value accruing to APIs and useful applications, not disappearing SaaS, and says Temporal is building integrations and Project Nexus for asynchronous tools. He closes by describing the CEO handoff from cofounder Max and urging founders to prioritize customer value, resilient margins, and focused growth.
章節
- 0:00Introduction: Temporal Resumes Failed Work and Processes Restaurant Orders Exactly Once
- 4:03Temporal’s Origin Story: Uber’s Three-Day Tip Retries and Loyalty-Point Recovery
- 11:14Why Agents Raise the Stakes: Multi-Region Failover and an Explosion of AI Apps
- 16:00Specialized Agents Need Durable RPC: From Claude Code Sandboxes to Hundreds of Agents
- 20:01Specialized Agents Need Durable RPC: OpenAI Codex Runs Background Tool Workflows
- 25:20Deep Research Agents: Recoverable Parallel Research and 150,000 Actions per Second
- 30:58Execution Histories as a Superpower: Auditable Agent Workflows and Tool Analytics
- 35:03Execution Histories as a Superpower: APIs, Rising App Traffic, and Temporal’s Integrations
- 39:04Minimal Viable Long-Running Agent Architecture: Sandboxes, Observability, and Enterprise Systems of Record
- 45:07Context Engineering at Scale: Real-Time Retrieval, Large Payloads, and Durable RPC
- 52:40Where Value Accrues: Five-Layer Cake, Abridge, Harvey, and Temporal’s CEO Transition
- 57:26Where Value Accrues: Temporal’s Post-2021 Resilience, Customer Focus, and R&D Raise
這是 Tier 1 公開摘要
每章重點、段落總結、心智圖由分享者控制是否公開。想看完整分析?自己提交一支。
同頻道的其他分析
To Regulate AI Effectively, Focus on How It’s Useda16z Deep DivesMartin Casado argues AI laws should target harmful uses, while regulatory uncertainty pushes US startups toward Chinese open-source models.
Mintlify and the Transition From Human Docs to Agent Infrastructurea16z Deep DivesMintlify grew from eight pivots and a two-day prototype into documentation infrastructure serving AI agents and 20 million monthly visitors.
Inferact: Building the Infrastructure That Runs Modern AIa16z Deep DivesInferact’s vLLM founders aim to build an open inference layer for models running across 400,000–500,000 GPUs.
How Palantir Scaled: Why the Best Software Is Built Backwardsa16z Deep DivesPalantir architect Akshay Krishnaswamy explains how field engineers turn customer pain into products and scale software backwards.
Braintrust CEO on Where Engineering Actually Matters in AIa16z Deep DivesBraintrust CEO Ankur Goyal says AI engineering should focus on evals and harnesses, while SQL beat Bash in his agent benchmark.
相關主題的分析
GPT-5.5 深入解析:為什麼從 Claude Code 跳到 Codex? | S2E55矽谷輕鬆談 Just Kidding TechGPT-5.5 長上下文準確率大幅提升,Kenji 實測 Codex 更順手,但提醒任務完成率仍須人工驗證。
AI 生產力的幻覺:從開發者到整個產業的自我感覺良好 | S2E60矽谷輕鬆談 Just Kidding TechMETR 實驗發現開發者用 AI 實際慢 19%,影片也提醒 Junior 可能因缺少訓練而出現能力斷層。
Peter Steinberger: "Fun Is Velocity"Y CombinatorPeter Steinberger says OpenClaw hit 4.7 million weekly downloads, but 9,500 configuration options and security work nearly overwhelmed him.
投資加密貨幣實測!3間交易所出入金使用心得(幣安_OKX_派網)|慢活夫妻慢活夫妻 George & DewiGeorge 與 Dewi 拆解加密貨幣入金、出金流程,並比較幣安、OKX 與派網的安全和功能。
AI Copilots Are a Dead End. Here's What Actually Works | Kavak CEOa16z Deep DivesKavak CEO Carlos García Ottati says AI agents now handle 90–95% of customer interactions after a year of flat growth during the transition.