AI Evaluation
522 episodes · Page 10 of 22
#3241: How to Write a Product Spec That Makes AI Find You the One
Stop typing three words into Google. The SPEC framework helps AI find exactly what you need.
#3171: How to Break an LLM's Bad Verbal Habits
Blacklists fail and regex inverts meaning. Here's what actually works to clean up AI writing tics.
#3170: Pharmacokinetics vs Neural Nets: Two Meanings of "Model
Two things called "models" that work completely differently — and why the confusion matters for patient safety.
#3157: Opus 4.8: What Actually Changed Under the Hood
Anthropic dropped Opus 4.8 with no fanfare. New training data, faster inference, and smarter refusals — here's what changed.
#3120: What Makes Agentic Search Tools Like Exa Actually Work?
Why swapping Google for Exa transformed our show's accuracy — and what agentic search does differently.
#3098: The Pilot with the Flashlight: Inside Aviation's Pre-Flight Walkaround
Why pilots still physically inspect planes before every flight — and what a 1979 crash taught us about trusting machines.
#3072: What Archival Actually Means in Your Pen
The AP seal isn't a durability guarantee. Here's what makes a marker truly archival.
#2950: Barley Beyond Soup: A Grain Guide
Pearl, pot, hulled, hulless — why barley labels matter for nutrition, cooking, and flavor.
#2936: Why AI Still Can't Really Teach You to Code
Code generators ship code. Real tutors build understanding. Why the gap is bigger than you think.
#2924: When Adding One Agent Breaks Everything
The math behind why your 100-agent pipeline fails 40% of the time — and what to do about it.
#2914: Can AI Read the Room? TTS Prosody Explained
Can TTS models truly infer emotion from text, or just mimic patterns? We break down the science of prosody.
#2909: The Reassurance Mirage: When Moderation Fails
How the EU Digital Services Act exposes a 30-to-1 gap in appeal success rates between platforms.
#2848: Can 100 Volunteers Let AI Govern Them for a Month?
An AI council of multiple models, a hundred volunteers, and a month of real municipal decisions. Here’s how you’d run the experiment.
#2847: How AI Could Transform Comparative Policy Analysis
Can AI agents do the work of a distributed think tank for cross-country policy learning?
#2800: The Two Meanings of Industrial Design
Industrial design is a profession. The "industrial look" is something else entirely. Here's where they split.
#2780: Building Self-Healing Agent Pipelines
How to build an agent that monitors and fixes other agents in production — without the hype.
#2754: Why Your Dictation Setup Might Be Wrong
Modern ASR is shockingly robust. The biggest predictor of accuracy? How well your audio matches its training data.
#2739: When Hoofbeats Are Zebras: How Doctors Learn to Think
How family doctors develop clinical judgment—pattern recognition, Bayesian reasoning, and the cognitive traps that lead to diagnostic errors.
#2736: Why AI Flagged Your Em Dash
Punctuation isn't a fixed system handed down by grammarians. It's a two-thousand-year story of contraction, invention, and now AI suspicion.
#2705: Your Brain Isn't a Hard Drive — What Actually Fits
Long-term memory isn't storage — it's a generative model. Here's where the brain/computer analogy actually holds up.
#2693: When AI Ignores Your Style Guide
Why your AI ignores formatting instructions and how to fix it with pipeline architecture, not model swaps.
#2688: Intelligent Frame Extraction for Multimodal AI
Use multimodal AI and smart frame extraction to turn a walk-through video into an actionable decluttering plan.
#2672: When a Startup Claims to Break the Quadratic Wall
A startup claims linear attention scaling at 12M tokens, beating GPT-5.5 on retrieval benchmarks.
#2665: Partner Certs vs Personal Certs: What Actually Matters
Solo operators face structural barriers in vendor partner programs. Here's how personal and partner certifications actually differ.