Stenobird

Episode

Claude Code for Finance + The Global Memory Shortage: Doug O'Laughlin, SemiAnalysis

Podcast
Latent Space: The AI Engineer Podcast
Published
Feb 24, 2026
Duration seconds
7453
Processing state
processed
Canonical source
https://www.latent.space/p/valuemule
Audio
https://api.substack.com/feed/podcast/189062462/654370814579f2242c870e8cb05b58a5.mp3
JSON
/v1/public/podcasts/latent-space-ai-engineer/episodes/claude-code-for-finance-the-global-memory-shortage-doug-o-laughlin-semianalysis
Markdown
/podcast/latent-space-ai-engineer/claude-code-for-finance-the-global-memory-shortage-doug-o-laughlin-semianalysis.md

Actions

  • POST https://stenobird.com/v1/public/podcasts/latent-space-ai-engineer/episodes/claude-code-for-finance-the-global-memory-shortage-doug-o-laughlin-semianalysis/transcription-requests
    Idempotently request low-priority transcript generation for this episode.
  • GET https://stenobird.com/podcast/latent-space-ai-engineer/claude-code-for-finance-the-global-memory-shortage-doug-o-laughlin-semianalysis.md
    Read the agent-friendly Markdown representation of this episode resource.

Summary

An exploration of how AI agents like Claude Code are transforming professional workflows from junior analyst tasks to high-level expertise. The discussion also dives into the physical constraints of the AI revolution, specifically the looming global memory shortage and semiconductor supply chain bottlenecks.

Topics

  • Claude Code
  • AI Engineering
  • Semiconductor Supply Chain
  • HBM Memory Shortage
  • Agentic Workflows
  • AI Economics
  • LLM Infrastructure
  • Software Automation

Highlights

  • Main idea: AI currently functions as a high-speed junior analyst, handling data gathering while humans provide the necessary meta-level expertise
  • Practical takeaway: Tools like Claude Code are rapidly automating GitHub repositories, with estimates suggesting 4% of GitHub is now written by AI
  • Failure mode: The 'Memory Mania'—a massive supply chain squeeze in HBM and memory is a primary bottleneck for scaling AI compute
  • Economic thesis: AI could act as a massive deflationary force, potentially challenging traditional metrics like GDP as information work becomes commoditized
  • Infrastructure reality: The future of AI scaling depends less on software and more on physical constraints like TSMC's capacity and optical interconnects

Chapters

  1. 0:00 AI as Junior Analyst: The role of LLMs in automating the 'grunt work' of information gathering and the importance of human expertise in the loop.
  2. 10:30 The Evolution of Research: Reflecting on the transition from macro-finance research to the current era of high-intensity semiconductor analysis.
  3. 29:00 The Software Engineering Frontier: Analyzing the massive shift in production traffic and the potential for AI to move beyond coding into broader business automation.
  4. 57:35 Agent Swarms and Automation: A reality check on agentic swarms, comparing them to traditional automation tools like Zapier.
  5. 1:07:00 The Economics of AI Scaling: Discussing the massive capital expenditures in AI and the potential for a 'Great Depression of AI' due to deflationary pressures.
  6. 1:35:30 The Memory and Supply Chain Squeeze: Deep dive into the HBM supply chain, the role of TSMC, and the critical importance of memory bandwidth and CXL.
  7. 1:54:25 Personal Reflections: A closing conversation on the value of intense physical experiences and self-mastery outside of the abstract digital world.