Opening Keynote

on YouTube ↗56m

" summary

Over half of Vercel deployments now originate from coding agents and token volume on the platform jumps from 2 trillion to 20 trillion tokens per month; the infrastructure updates are concrete: Fluid Compute 2.0 permits 30 minute functions, 5 GB uploads, Docker images on microVMs and active CPU billing (~$0.15 per active CPU hour). The agent stack ships AI SDK 7, AI Gateway, Workflow SDK, Sandbox, Connect and Chat SDK, and the demo of Eve scaffolds, connects Linear and Salesforce, runs sandboxed tool calls with human approvals and deploys to Slack; 1,500 teams and 40,000 agents are already active and 100+ agents run internally in production.

AI-written summary · about

" chapters (7)

1. 0:00The agentic web
2. 8:44Fluid Compute 2.0, Docker, and Vercel Services
3. 17:12The Agent Stack: AI SDK 7, AI Gateway, Workflow SDK, Sandbox, Connect, and Chat SDK
4. 29:41Introducing eve + live demo
5. 38:05Vercel Agent
6. 41:16100+ agents in production and Vercel for Enterprise Apps and Agents
7. 53:54Wrap-up

" related talks

• 54mKeynoteTom Occhino, Guillermo Rauch, Malte Ubl, Jeanne DeWitt Grosser, Hedi Zandi, Shai Dara · Vercel Ship 26 London
• 50mOpening KeynoteGuillermo Rauch, Malte Ubl, Pranay Prakash, Nico Albanese, Aparna Sinha · Ship AI 2025
• 27mCTO Panel: Agents in productionVercel Ship 26 London
• 19m1,200 analytics requests a day: How we built Vercel's data agentAbhi Sivasailam · Vercel Ship 26 London
• 34mRethinking the future of digital transactionsJeanne DeWitt Grosser, Kate Jensen, Arjun Rajeswaran, Jeff Weinstein · Ship AI 2025
• 31mFireside ChatGuillermo Rauch, Harry Stebbings · Vercel Ship 26 London
• 16mSoftware that ships itself: self-driving appsMalavika Balachandran Tadeusz · Vercel Ship 26 London
• 30mAccelerate AI to prodANDREW BAIRD · Ship AI 2025