Broadcast Archives

SR-008 | December 24, 2025 @ 9:00 PM EST

Syntax, Semantics, and the Shape of Collaboration: Designing Languages for Human and Machine

Guest

Chris Lattner (Creator of LLVM and Swift, Modular AI)

Explored how programming language design should evolve for AI collaboration, examining the balance between human ergonomics and machine comprehension. Discussed explicit structure, type systems, error handling patterns, verbosity versus clarity, domain-specific languages, formal specifications, readability standards, and tooling integration. Emphasized that good language design principles—clarity, consistency, safety—serve both human and machine readers, suggesting evolutionary refinement rather than revolutionary redesign.

SR-007 | December 23, 2025 @ 9:00 PM EST

Attack Surfaces and Automated Vulnerabilities: Security in the Age of Autonomous Code

Guest

Alex Stamos (Cybersecurity Expert, Former Facebook CSO)

Explored security implications of AI coding tools, examining training data poisoning risks, context manipulation through malicious comments, execution sandboxing challenges, prompt injection vulnerabilities, cloud service attack vectors, secrets exposure, verification of AI-generated security code, and systematic vulnerability introduction at scale. Emphasized that AI tools expand attack surfaces while requiring human security expertise to remain critical for safe deployment.

SR-006 | December 22, 2025 @ 9:00 PM EST

The Bounded Window: Attention, Architecture, and the Limits of Local Understanding

Guest

Andrej Karpathy (AI Researcher, Former Tesla and OpenAI)

Examined fundamental constraints of context windows in AI coding tools, exploring how transformers process limited text spans, the impossibility of full-project understanding, retrieval strategies and their trade-offs, attention mechanisms and pattern matching versus reasoning, architectural implications for codebase design, and why partial context makes testing and modularity more critical. Emphasized that context limits are fundamental constraints requiring good engineering practices rather than technical solutions.

SR-005 | December 21, 2025 @ 9:00 PM EST

The Paradox of Automated Productivity: Economics When Code Is Cheap

Guests

Patrick Collison (CEO, Stripe)
Dr. Erik Brynjolfsson (Economist, Stanford Digital Economy Lab)

Explored economic implications of AI-assisted development, examining productivity measurement challenges, the gap between output metrics and value creation, complementarity effects that amplify existing skill differences, organizational bottlenecks beyond coding speed, technical debt accumulation risks, labor market effects, and the distinction between short-term velocity and long-term lifecycle costs. Emphasized that true productivity requires organizational adaptation, not just faster coding.

SR-004 | December 20, 2025 @ 9:00 PM EST

Reviewing the Unwritten: Accountability in the Age of Algorithmic Authorship

Guest

Jessie Frazelle (Engineer, Former Docker and Google)

Examined how code review practices must adapt for AI-generated contributions, exploring the shift from catching implementation errors to verifying specification understanding. Discussed AI code signatures, the loss of knowledge transfer in reviews, ownership and accountability questions, review standards for machine-generated work, documentation requirements, and the need for explicit human responsibility. Emphasized that AI tools increase rather than diminish reviewer accountability.

SR-003 | December 19, 2025 @ 9:00 PM EST

The Discipline of Design: TDD and the Limits of Mechanical Virtue

Guest

Kent Beck (Software Engineer, Creator of Extreme Programming)

Explored whether AI agents can meaningfully practice test-driven development, examining the distinction between mechanical execution and design discovery. Discussed API design through tests, refactoring capabilities, test quality challenges, the rhythm of red-green-refactor cycles, and comparisons between AI and human pair programming. Emphasized that TDD's value lies in shaping human thinking, not just producing tests, and proposed AI as amplification rather than substitution for developer competence.

SR-002 | December 18, 2025 @ 9:00 PM EST

Planning, Parsing, and the Problem of Partial Understanding

Guest

Simon Willison (Independent Researcher, Creator of Datasette)

Examined Claude Code's internal representations, multi-file editing architecture, and planning mechanisms. Discussed dependency tracking failures, the critical role of test coverage, git integration, and error handling loops. Explored how the system balances structural and semantic code understanding, the importance of human plan review, and future possibilities for formal verification integration.

SR-001 | December 17, 2025 @ 9:00 PM EST

Agentic Autonomy and the Architecture of Assistance

Guest

Dario Amodei (CEO, Anthropic)

Explored Claude Code's project-level reasoning capabilities, multi-file editing, and pattern recognition. Discussed the shift from code generation to verification, challenges of context limitations, responsibility for AI-generated code, and performance degradation for uncommon languages. Examined how autonomous coding tools reshape developer workflows and the essential role of human review.