AMD just shipped ROCm 10 and took ROCm.AI to general availability. Installing the stack is now a sentence you type into Claude Code or Codex, an open source agent system called Hyperloom rewrites your kernels while you sleep, and AMD claims a 3.3x average inference lift and 2.4x training lift over ROCm 7 on the same hardware.

Anush Elangovan, Corporate VP of AI Software at AMD, is back for his third appearance to walk through what actually changed, why he still runs his agents with permissions skipped, and where the bottleneck moved once writing code stopped being the hard part.

His answer: the infrastructure we built for humans does not survive contact with agent swarms.

We cover:

Why the entry point to ROCm is now natural language instead of a docs page and a curl command
How Hyperloom turns kernel optimization into an overnight search problem, including the pass where AMD optimized 14,000 models
The success metric Anush uses for ROCm.AI: someone who cannot spell ROCm getting an LLM served
Why a fully open source corpus keeps AMD Skills from going stale as frontier models turn over
Agent swarms as African wild dogs, and the security posture that breaks when 20 years of attacks run in 30 seconds
Who is liable when an agent takes down a power grid
Why the last mile of AI will move more value than the next frontier model

Chapters:

(0:25) A decade of ROCm, now agent native
(3:03) What agentic ROCm looks like in practice
(6:17) Installing ROCm then versus now
(9:14) An order of magnitude more CI across every framework
(10:44) Anush's workflow: skip permissions, deploy in one sentence
(12:21) Speed is the moat
(15:00) Success is a stranger who cannot spell ROCm serving an LLM
(17:09) Keeping agent skills from going stale
(21:38) Co-designing kernels with the frontier labs
(24:06) Hyperloom, GEAK, and 14,000 models in one pass
(26:45) Agent swarms and the liability question
(32:21) African wild dogs and the new security posture
(36:03) 3.3x over ROCm 7, and where the bottleneck moved
(37:36) Where enterprises hit walls in production
(40:24) Why coding was the right reward function for AI
(44:42) Which industries get the next software scale unlock
(47:02) The last mile of AI
(50:31) Closing thoughts

Connect with Anush Elangovan:

LinkedIn: https://www.linkedin.com/in/anushelangovan/
Twitter/X: https://x.com/AnushElangovan
ROCm.AI: https://rocm.ai
AMD AI blog: https://www.amd.com/en/blogs/by-author/anush-elangovan.html
AMD AI Developer Program: https://www.amd.com/en/developer/ai-dev-program.html

Connect with Chain of Thought host Conor Bronsdon:

Newsletter: https://newsletter.chainofthought.show/
Twitter/X: https://x.com/ConorBronsdon
LinkedIn: https://www.linkedin.com/in/conorbronsdon/
YouTube: https://www.youtube.com/@ConorBronsdon

More episodes: https://chainofthought.show

Thanks to Svix, presenting sponsor of season four of Chain of Thought. Svix delivers billions of reliable webhooks for startups and the Fortune 500. Get started at https://link.svix.com/cot. Qualified startups get $12,000 in credits, and YC companies get $50,000.

Thanks to Walrus, presenting sponsor of season four of Chain of Thought. Walrus Memory gives AI agents portable, verifiable memory that carries context across apps, sessions, and other agents. Get started at https://walrus.xyz/cot

AI Agents Are Outgrowing Human Infrastructure | AMD's Anush Elangovan

Chain of Thought | AI Agents, Infrastructure & Engineering

AI Agents Are Outgrowing Human Infrastructure | AMD's Anush Elangovan