What Topics Does the Chain of Thought Podcast Cover?

Updated

The Chain of Thought Podcast covers AI infrastructure, AI agents, and AI evaluation and reliability, with recurring subtopics including enterprise AI, AI security, multi-agent systems, RAG and retrieval, AI hardware, AI observability, AI coding, agent memory, multimodal AI, and context management. It is hosted by Conor Bronsdon and features guests from companies such as NVIDIA, Google DeepMind, and AMD. If your interest is in how AI systems are actually built and operated in production—rather than general AI news—the topic range is likely to match.

Core themes

The show describes itself as "conversations with the people building AI agents and infrastructure," aimed at engineers, founders, and operators shipping AI. Three themes anchor most episodes:

  • AI infrastructure — the systems, hardware, and serving layers that AI products run on.
  • AI agents — how agents are designed, given context and memory, and deployed.
  • AI evaluation and reliability — how teams measure whether AI systems work.

Subtopics you can browse

The site groups episodes and explainers under a topic navigation, which is the fastest way to check whether your area is covered:

Topic What it points to
AI Agents Agent design, guardrails, deployment
AI Evaluation & Reliability Evals, LLM-as-a-judge vs. human review
Enterprise AI Why enterprise AI projects do or don't show ROI
Open Source AI Open model and tooling discussions
RAG & Retrieval Retrieval-augmented generation
AI Hardware Compute and infrastructure hardware
AI Observability Monitoring AI systems in production
AI Coding AI-assisted software development
Agent Memory How agents retain and use context
Multimodal AI Models handling more than text
Context Management Supplying the right context to models

Two content formats

The podcast publishes in two forms, and they serve different needs:

  1. Episodes — long-form guest interviews. Recent examples include an episode on switching models without starting over (Walrus Protocol), one on AI engineering metrics (Sonar), and one on why context alone isn't enough for enterprise AI agents (WisdomAI). Episodes run roughly 56–83 minutes based on the listed durations.
  2. Explainers — short "plain answers to the questions builders ask," each linked to guests who have dealt with the issue in production. Examples include whether to build a single agent or a multi-agent system, how to give an agent identity and permissions, and what AI evals are.

How to decide if it fits

  • Choose it if you want practitioner-level discussion of AI infrastructure, agents, and evals, and you value guest experience from production deployments.
  • Look elsewhere if you want beginner-level AI introductions or broad industry news rather than build-and-operate detail.

To check fit quickly, start with the explainers page for a topic you care about, then move to the full episode list. The show is available on YouTube, Spotify, Apple Podcasts, and RSS, and there is a free email newsletter with new episodes and essays.