Show Notes
Samar Abbas, co-founder and CEO of Temporal, joins Claire Zau and Lightspeed partner Anoushka Vaswani to break down a $550 million Series E at a $12.55 billion valuation, seven months after a round that valued the company at $5 billion. Temporal is an open source platform for durable execution, the guarantee that code runs to completion no matter what fails underneath it. Samar explains why that matters more in the agent era. The longer an agent runs and the more tools it touches, the more the chances of failure compound, and the real world is full of cases where a retry means refunding a customer twice. Anoushka makes the case that reliability and long running execution are the two things every useful agent needs, which puts Temporal in the path of token flow as agents do more work. Samar walks through the subscription workflow that can take “four lines of code” instead of a stack of queues and timers, and the nearly twenty year arc from Simple Workflow at Amazon through Cadence at Uber. They close on Temporal Nexus and what it would take to extend the same guarantees across company boundaries, so one business's agent can reliably call another's. Chapters: 0:00 Temporal’s $550M Series E and $12.55B Valuation 0:45 What Temporal Does and How It Works 3:00 How Temporal Can Solve a Critical AI Infrastructure Problem 4:16 Where AI Models Fail in Real-World Execution 7:04 How Temporal Changes the Developer Experience 8:43 Temporal’s “Aha Moment” for AI Teams 9:43 Interesting Temporal Use Cases 11:44 From Simple Workflow at Amazon to Cadence at Uber 14:11 Temporal's Partnership with Lightspeed 17:59 How Temporal Will Use the New Funding 19:13 The Future of Agents Communicating With Each Other 22:38 What Will Become Obvious About AI in the Future
More from Lightwork
Meta and Google are making their next moves in the race for frontier AI, Apple is pushing deeper into the smart home, and SpaceX wants to become a major player in mobile. Claire and Josh discuss Meta’s Muse Wonder and Wonder Full models, including the company’s focus on active memory, persistent agents, and keeping its most capable model inside the Muse ecosystem. They also examine Google’s Gemini 4 Argon and its strengths across professional workflows, vision, coding, spatial reasoning, and long-running tasks. Then, they look at Apple’s HomeView and why privacy, personalization, and ambient intelligence could give Apple an opening in the AI-powered home. Plus, they unpack what SpaceX would actually need to become a major U.S. wireless carrier and how a company that controls satellites, spectrum, and broadband could reshape the telecom market. Chapters: 0:00 Welcome to Lightwork 1:04 Meta Muse Wonder vs. Google Gemini 4 5:00 Why AI Agents Need Memory and Focus 6:23 Gemini’s Advanced Reasoning and Task Automation 8:37 Apple HomeView: Siri Comes Home 11:33 AI Privacy and Smart Homes 16:52 SpaceX’s Plans To Be a Carrier 22:12 Zaria $5.2M Seed Round Led By Lightspeed 22:45 Zenzap Comes Out of Stealth Stay in touch: www.lsvp.com X: https://twitter.com/lightspeedvpLinkedIn: https://www.linkedin.com/company/lightspeed-venture-partners/Instagram: https://www.instagram.com/lightspeed/ The content here does not constitute tax, legal, business or investment advice or an offer to provide such advice, should not be construed as advocating the purchase or sale of any security or investment or a recommendation of any company, and is not an offer, or solicitation of an offer, for the purchase or sale of any security or investment product. The views expressed by our guests do not necessarily represent the views of Lightspeed. For more details please see lsvp.com/legal.
Reflection co-founder Misha Laskin sits down with Claire Zau in Washington, D.C. to explain why AI labs need a presence in policy conversations and where the open weights debate stands right now. Misha compares closed and open models to renting and owning a home, and says companies that started by renting intelligence are now moving toward ownership. He describes a shift on AI gateways over the past six months from roughly 70 percent closed tokens to a majority of open ones. They also cover some of Misha’s arguments: why China's labs have led on open models, why AI startups want control over cost, compute and data, and why that open source is the safer way to deploy AI. He makes the case through the history of strong encryption and Linus's law, and through the recent incident involving OpenAI and Hugging Face. Stay in touch: www.lsvp.comX: / lightspeedvp LinkedIn: / lightspeed-venture-partners Instagram: https://www.instagram.com/lightspeed/...The content here does not constitute tax, legal, business or investment advice or an offer to provide such advice, should not be construed as advocating the purchase or sale of any security or investment or a recommendation of any company, and is not an offer, or solicitation of an offer, for the purchase or sale of any security or investment product. The views expressed by our guests do not necessarily represent the views of Lightspeed. For more details please see lsvp.com/legal.Chapters: 00:00 Why AI Labs Need a Presence in DC01:15 The Open Weights Fight Explained02:09 Closed vs. Open Models: The Real Estate Analogy03:16 Why China Is Leading Open Models05:05 Why Developers and Startups Are Adopting Open Models06:45 AI Safety and the Open Source Debate08:19 Linus’s Law Applied to AI Safety09:26 Why Open Models Are Actually Safer10:44 Open Models in Practice: The Hugging Face IncidentvShifting From Renting AI to Owning It • Shifting From Renting AI to Owning It
Josh Machiz is joined by Shaurya Aggarwal, a partner on Lightspeed's growth team, for the week the West showed up to the open model fight. Reflection unveiled Beam on Monday, and Mistral unveiled Large 4, better known as Le Chonk, on Tuesday. Shaurya explains what open weights actually buy you, why both companies are competing on intelligence per dollar instead of chasing the frontier, and he breaks down mixture of experts in a way that makes the huge parameter counts finally make sense. Only about 5% of Beam wakes up for any given token. Lightspeed led Mistral's inception round and has invested in every round since, and co-led Reflection's Series A.Then DoorDash, which showed a delivery drone it designed itself plus an AI agent you can text to order. Shaurya's take is that the marketplace is the moat, and building its own agent is offense and defense at once against the assistants people already use. On that note, Instinct can now join your group chat, and Meta and Sierra announced an open standard for how personal agents work with businesses. Finally Oura, which called off its Nasdaq IPO hours before pricing despite heavy demand, and what that says about late stage growth investing. Plus updates from RobCo and Avarra. Lightspeed led RobCo's Series B, co-led its Series C and participated in this round, and led Avarra's seed and participated in its Series A.Chapters:00:00 Open Models, Drone Delivery, and AI Agents01:08 Mistral's Le Chonk Meme and Naming Story02:33 What Open-Weight Models Mean for the AI Race03:56 When to Choose Open Models Over Frontier AI05:30 Why Open Models Compete on Cost Efficiency06:35 Mixture of Experts Explained07:32 Is Reflection the DeepSeek of the West?07:58 Why DoorDash Is Building Its Own Delivery Drones09:38 DoorDash's Competitive Edge 12:28 DoorDash's Moat Against AI Agents14:34 Personal AI Agents in Group Chats16:08 How AI Agents Could Change Consumer Spending17:26 Meta, Sierra, and the Future of Personal AI Agents19:56 Oura Postpones Its IPO22:13 Lightspeed UniverseStay in touch: www.lsvp.comX: / lightspeedvp LinkedIn: / lightspeed-venture-partners Instagram: / lightspeedventurepartners The content here does not constitute tax, legal, business or investment advice or an offer to provide such advice, should not be construed as advocating the purchase or sale of any security or investment or a recommendation of any company, and is not an offer, or solicitation of an offer, for the purchase or sale of any security or investment product. The views expressed by our guests do not necessarily represent the views of Lightspeed. For more details please see lsvp.com/legal.Reflection Beam, Mistral Le Chonk & the Drone That Delivers Your Chipotle | Lightwork • Reflection Beam, Mistral Le Chonk & the Dr...
Robotics had a big summer, and it is genuinely hard to tell who is doing what. Claire Zau met with Lightspeed partner Amber Yang in front of a whiteboard to draw the whole market on two axes. How general is the model, and how general is the body. Yang explains what actually changed at the model layer, why robots cannot just be trained on the internet the way language models were, and why she thinks humanoids are not about to get cheap. Roughly 40 to 60 percent of a humanoid build is actuators, and that supply chain runs through China. Then she fills in the map. Dexterity, Tacta and Reflex Robotics building their own models for their own hardware. Ultra skipping training entirely and fine-tuning on Physical Intelligence. Skild and Physical Intelligence chasing one brain for many bodies. Sunday Robotics, 1X and Figure on the humanoid side. She also makes the case for what she calls DeployCos, companies that take someone else's model into places like a restaurant or a nursing home without the customer ever knowing what is under the hood. Her closing point is the one we believe worth sitting with. Robotics still has no shared benchmark, so every company grades its own homework, and in a field this tied to safety that gap may be an opportunity. Stay in touch: www.lsvp.comX: https://twitter.com/lightspeedvpLinkedIn: https://www.linkedin.com/company/lightspeed-venture-partners/Instagram: https://www.instagram.com/lightspeed/The content here does not constitute tax, legal, business or investment advice or an offer to provide such advice, should not be construed as advocating the purchase or sale of any security or investment or a recommendation of any company, and is not an offer, or solicitation of an offer, for the purchase or sale of any security or investment product. The views expressed by our guests do not necessarily represent the views of Lightspeed. For more details please see lsvp.com/legal.
Claire Zau and Josh Machiz were both in DC this week, stuck in traffic for a White House lunch they didn't know was happening until they saw the seating chart. About twenty executives were in the room, and what came out was the White House Accord on Superintelligence, a one page document covering internal monitoring, verification teams, an independent auditor, and a board committee. Alongside it, an executive order renaming artificial intelligence to superintelligence across federal documents. Claire breaks down what the accord actually commits anyone to, and why a lunch is not a law.Then the government's own product launch. America.gov puts one chatbot in front of roughly 29,000 government sites, the ones about 40 million people visit daily, and Josh has been testing it. From there, OpenAI's DevDay and more than 20 launches, led by Dots, an always on agent with its own computer that you can voice call, plus GPT-6.1 Sol at a fifth of Astra's token price. Claire and Josh get into what Josh calls the Tamagotchification of tech, and whether anyone wants four of these assistants. Finally Starship, which reached orbit for the first time after 13 suborbital tests and deployed 26 V3 Starlink satellites, having lost an engine on the way up. Plus updates from ElevenLabs and Modus. Lightspeed joined ElevenLabs' Series D and led Modus' seed and Series A.Stay in touch: www.lsvp.comX: https://twitter.com/lightspeedvpLinkedIn: https://www.linkedin.com/company/lightspeed-venture-partners/Instagram: https://www.instagram.com/lightspeedventurepartners/Chapters: 00:00 Welcome Back to Lightwork!01:15 Stuck in DC Traffic for a White House Lunch01:52 Who Was in the Room03:32 The White House Accord on Superintelligence04:48 Trump Renames AI to Superintelligence08:01 America.gov: One Chatbot for 29,000 Government Sites10:03 Josh Tests the Chatbot12:20 OpenAI DevDay: Introducing Dots15:05 The Tamagotchification of Tech19:37 Do You Really Want Four AI Assistants?19:47 GPT-6.1 Sol: Frontier Intelligence at a Fifth of the Cost24:20 Starship Reaches Orbit After 13 Suborbital Tests29:36 The First Commercial Starship Flight31:05 ElevenLabs V4 and Modus Audit BenchThe content here does not constitute tax, legal, business or investment advice or an offer to provide such advice, should not be construed as advocating the purchase or sale of any security or investment or a recommendation of any company, and is not an offer, or solicitation of an offer, for the purchase or sale of any security or investment product. The views expressed by our guests do not necessarily represent the views of Lightspeed. For more details please see lsvp.com/legal.