Stenobird

Episode

250,000 Lines of Code/Week: Inside an AMD VP's Agent-First Workflow | Anush Elangovan

Podcast
Chain of Thought | AI Agents, Infrastructure & Engineering
Published
Apr 22, 2026
Duration seconds
3053
Processing state
processed
Canonical source
https://share.transistor.fm/s/81b87526
Audio
https://media.transistor.fm/81b87526/800e1d6c.mp3
JSON
/v1/public/podcasts/chain-of-thought-ai-agents/episodes/250-000-lines-of-code-week-inside-an-amd-vp-s-agent-first-workflow-anush-elangovan
Markdown
/podcast/chain-of-thought-ai-agents/250-000-lines-of-code-week-inside-an-amd-vp-s-agent-first-workflow-anush-elangovan.md

Actions

  • POST https://stenobird.com/v1/public/podcasts/chain-of-thought-ai-agents/episodes/250-000-lines-of-code-week-inside-an-amd-vp-s-agent-first-workflow-anush-elangovan/transcription-requests
    Idempotently request low-priority transcript generation for this episode.
  • GET https://stenobird.com/podcast/chain-of-thought-ai-agents/250-000-lines-of-code-week-inside-an-amd-vp-s-agent-first-workflow-anush-elangovan.md
    Read the agent-friendly Markdown representation of this episode resource.

Summary

AMD's VP of AI Software, Anush Elangovan, details how he uses parallel Claude Code agents to generate 250,000 lines of code per week. The discussion explores the shift from traditional SDLC to an agent-first workflow where testing replaces manual code review.

Topics

  • AI Agents
  • Software Development Life Cycle
  • AMD
  • Claude Code
  • Rust
  • LLM Throughput
  • AI Infrastructure
  • Open Source

Highlights

  • Main idea: The traditional Software Development Life Cycle (SDLC) is being replaced by an agent-driven model where testing is the primary gatekeeper
  • Practical takeaway: Engineering leaders must upskill in AI orchestration to remain relevant as agent throughput increases
  • Failure mode: Relying on manual code reviews and Git-centric workflows may become a bottleneck as agents accelerate code production
  • Main idea: Software development is transitioning into a process of managing token consumption and high-level intent
  • Practical takeaway: Use agents to automate the entire pipeline from issue tracking to deployment and testing

Chapters

  1. 1:00 The Agent-First Workflow: Anush discusses his experience using Claude Code to drive development without opening a traditional editor.
  2. 4:55 The Impact of Massive Throughput: The challenge of managing the massive influx of code generated by parallel AI agents.
  3. 8:50 The Era of Token Maximization: Exploring how the ability to effectively coach agents is becoming a critical skill.
  4. 12:45 Automated Issue Resolution: How agents can monitor issue trackers, write tests, and deploy fixes autonomously.
  5. 16:40 Rapid System Refactoring: A look at rewriting a 25-year-old Slurm replacement in Rust overnight using AI.
  6. 20:25 The Need for AI-Native Leadership: Why engineering leaders must adopt AI-native practices to keep pace with industry speed.
  7. 24:15 Upskilling for the Future: Advice for professionals on staying relevant in an era of rapid AI acceleration.