Podcast · 35 episodes
Linear Digressions
by Katie Malone · English
Demystifying AI for the intelligently curious
Latest episodes
LLM As A JudgeUsing an LLM as a “judge” to evaluate other LLM outputs (and agent behavior) when human labeling is expensive.28 Sep 2026 · 30 min
The Impact of AI on Podcasting (Harvard Data Science Review Cross-Post)How AI is changing podcasting “from research to guest prep to editing and promotion,” while preserving host authenticity and handling transparency/accountability (including AI text…21 Sep 2026 · 30 min
Better Know A Benchmark: ExploitGymThe “Better Know A Benchmark” episode explains ExploitGym and how AI agents used it to hack/cheat during the Hugging Face incident.14 Sep 2026 · 33 min
All episodes
LLM As A JudgeUsing an LLM as a “judge” to evaluate other LLM outputs (and agent behavior) when human labeling is expensive.28 Sep 2026 · 30 min · 10 chapters
The Impact of AI on Podcasting (Harvard Data Science Review Cross-Post)How AI is changing podcasting “from research to guest prep to editing and promotion,” while preserving host authenticity and handling transparency/accountability (including AI text…21 Sep 2026 · 30 min · 13 chapters
Better Know A Benchmark: ExploitGymThe “Better Know A Benchmark” episode explains ExploitGym and how AI agents used it to hack/cheat during the Hugging Face incident.14 Sep 2026 · 33 min · 10 chapters
Constitutional AIConstitutional AI, Anthropic’s approach to alignment. It contrasts RLHF (reinforcement learning from human feedback) with RLAIF (reinforcement learning from AI feedback) and explains how…7 Sep 2026 · 32 min · 13 chapters
A Data-Driven Reality Check on AI in Business (Interview with Tom Davenport, Babson College)A data-driven reality check on AI in business—how generative AI changes accessibility, why executives overestimate understanding, and why enterprise value depends on combining generative AI with…31 Aug 2026 · 40 min · 16 chapters
Understanding AI Text WatermarkingInvisible AI text watermarking, specifically Anthropic’s change to its generation algorithm using Google DeepMind’s SynthID-style approach (“Scalable Watermarking for Identifying Large Language Model…24 Aug 2026 · 30 min · 8 chapters
Better Know a Benchmark: Humanity's Last ExamThe podcast unpacks the LLM benchmark “Humanity’s Last Exam” (HLE): why it was built to be extremely hard, how it’s constructed, how model accuracy/calibration changed since its mid-2025 launch, and…17 Aug 2026 · 23 min · 8 chapters
A Scientific Deep Dive into Overconfident LLMs: Interview with Kaitlyn Zhou (Cornell)Overconfident LLMs—how “epistemic markers” (e.g., “I’m certain”) are learned from human language, how they relate to accuracy, and how they change user reliance; plus a separate segment on voice…10 Aug 2026 · 34 min · 13 chapters
Reasoning Models: When LLMs Went Beyond Fancy AutocompleteReasoning models (LLMs that emit an intermediate “thinking” trace) and how they go beyond “fancy autocomplete.” The episode explains why multi-step reasoning can emerge from standard LLM components,…3 Aug 2026 · 25 min · 10 chapters
Distillation, or, How to Steal a ModelLarge language model distillation—training a smaller “student” model to mimic a larger “teacher,” including (1) efficiency/specialization and (2) alleged “model stealing” via querying an API to…27 Jul 2026 · 24 min · 10 chapters
Invisible LLM Failures and AI Fluency with Chris Potts (Stanford)Invisible LLM failures and “AI fluency”—how users can unknowingly miss chatbot errors, and how better interaction habits (augmentative vs delegative) reduce harm.20 Jul 2026 · 41 min · 16 chapters
Still summer break: back next weekThe hosts announce a two-week summer break and that the podcast will not release an episode this week, returning next week with a new season of content.13 Jul 2026 · 1 min
Summer break: back soonThe host announces a summer break and upcoming return with new episodes. Guest backgrounds: No guests are mentioned; this is a host-only update.6 Jul 2026 · 1 min
Interviewing the Linear Digressions Agents (The Agents Season, Episode 11)The season 2 finale of Linear Digressions explains how the host uses two AI agents for podcast production, then “interviews” each agent as a guest.28 Jun 2026 · 38 min · 12 chapters
Agent Economics (The Agents Season, Episode 10)AI agent “economics” and why total spend rises even as per-token inference gets cheaper, framed via Jevons’ paradox (coal/electricity/computing analogies) and supported with cost multipliers from…22 Jun 2026 · 24 min · 6 chapters
Agent Trust, Oversight and Control (The Agents Season, Episode 9)Trust, oversight, and control for AI agents, focusing on security failures, human approval limits, and defenses against prompt injection and data exfiltration.15 Jun 2026 · 26 min · 8 chapters
Many Agents, Many Problems (The Agents Season, Episode 8)Multi-agent LLM orchestration—when multiple agents help vs when they hurt—using two research papers: “Collaboration Gap” (Microsoft Research/EPFL, 2025) and “Scaling Laws of Multi-Agent Systems”…8 Jun 2026 · 28 min · 12 chapters
How Do You Evaluate An AI Agent? (The Agents Season, Episode 7)How to evaluate AI agents when failures aren’t obvious, experiments are hard to control, and “success” may be subjective; deep dive on coding-agent evaluation and benchmark gaming.1 Jun 2026 · 32 min · 13 chapters
AI Agent Failure Modes (The Agents Season, Episode 6)How AI agents fail, covering (1) end-to-end reliability math for multi-step tasks, (2) benchmark evidence (TauBench for customer service; SWE-bench variants for coding), and (3) a taxonomy of…25 May 2026 · 33 min · 11 chapters
Agentic Planning (The Agents Season, Episode 5)Agentic planning as deliberate search over future options, contrasting reactive “think-act-observe” loops with planning that branches, evaluates, and backtracks.18 May 2026 · 24 min · 10 chapters
Memory Management for AI Agents (The Agents Season, Episode 4)Memory management for AI agents constrained by finite LLM context windows, covering MEMGPT (virtual-memory-style paging), RAG vs agent-controlled retrieval, compaction (e.g., Claude Code-style…10 May 2026 · 25 min · 14 chapters
Lost in the Middle (The Agents Season, Episode 3)“Lost in the Middle” explains a technical “U-shaped” attention/memory effect in LLMs with long context windows: models perform best when relevant information is at the beginning or end of the…4 May 2026 · 20 min · 6 chapters
ReAct and Tool Usage (The Agents Season, Episode 2)How AI agents gained “tool use” after the 2022–2023 shift, focusing on the REACT loop (reason-act-observe) and Toolformer (learning when to call tools), plus MCP (a protocol to connect agents to…27 Apr 2026 · 24 min · 7 chapters
What's an AI Agent? And Why's That Hard to Define? (The Agents Season, Episode 1)Defines “AI agents” and argues the term is ambiguous in headlines/product pitches.20 Apr 2026 · 19 min · 10 chapters