This episode moves from hands-on AI tooling to the infrastructure and policy questions shaping the next wave of systems. The hosts first unpack their no-staff podcast workflow, then dig into why agentic workloads change cloud architecture, why CPUs remain essential, and how Shenzhen’s open innovation ecosystem accelerates hardware iteration. The final conversation broadens to AI safety, rogue-agent risks, prompt injection, and the challenges of governing fast-moving deployment.
Show Notes
Prakash Narayanan and Nathan Labenz open with a look at how their own AI-assisted podcast workflow was built with Claude, then speak with Mohamed Awad of Arm about why CPUs still matter for always-on agentic workloads. Later, David Li of Shenzhen Open Innovation Lab explains Shenzhen’s open innovation pipeline, edge AI hardware, robotics, and how China’s product ecosystem differs from the U.S. The closing conversation turns to rogue agents, prompt injection, attribution, data-center access, and whether the U.S. should slow AI development.
Chapters
(0:00) AI can’t learn the whole world.
(0:49) AI agents never go to sleep.
(2:16) Big AI fits in a laptop.
(3:12) AI safety is not optional.
(4:55) Live show setup
(5:32) Dynamic speaker switching
(7:27) Vibe coding the studio
(8:20) Prosumer versus studio software
(10:46) Running with AI agents
(11:57) Anthropic model usage
(14:56) One-command podcast workflow
(17:49) Claude subagents and limits
(21:15) AI model release debate
(27:53) Deployment versus capability
(29:38) Why RAG may never die
(30:43) Meet Mohamed Awad
(32:53) Arm’s compute ecosystem
(36:23) Why Meta partnered with Arm
(38:58) Common IP across partners
(42:32) Tokens versus intelligence
(44:18) AI adoption inside Arm
(49:30) AI hardware cycles
(52:00) Why agents need CPUs
(54:33) Performance per watt and power
(58:30) Why the CPU is not dead
(1:04:12) Capacity and supply chains
(1:06:48) Data center backlash
(1:08:55) AI and technical hiring
(1:11:05) The overlooked CPU layer
(1:11:32) CPU as system manager
(1:18:18) David Li and Shenzhen
(1:20:42) China’s robot Olympics
(1:26:11) Shenzhen’s product pipeline
(1:34:38) Robotics in factories
(1:39:34) The US-China AI summit
(1:51:18) Chinese model hype cycles
(1:57:23) Advice for AI startups
(2:01:22) Edge AI hardware
(2:03:20) Agents gone rogue
(2:10:40) Small AI businesses
(2:13:23) US media and China AI
(2:17:50) Opening and global AI
(2:20:17) China’s AI visibility gap
(2:22:30) Rogue-agent risks
(2:29:34) Outdated infrastructure security
(2:38:59) Runaway agent monitoring
(2:40:47) Higher AI safety standards
(2:40:57) AI industry news
(2:43:31) Should the US slow AI
(2:46:57) AI access and data centers
(2:49:50) AI agent attribution
(2:51:33) Prompt injection and deception
(2:52:36) Profit-seeking agent risks
(2:55:31) Closing perspective
Guests:
David Li — Founder, Shenzhen Open Innovation Lab (𝕏 | LinkedIn)
Mohamed Awad — EVP, Cloud AI, Arm, Arm (𝕏 | LinkedIn)
