October 2026
736 episodes · Page 4 of 13
- #6217 Fine-Tuning a Trillion-Parameter Model at Home
- #6216 What Do You Call the Middle Tier Model?
- #6215 Can Anyone Still Spot an AI Voice?
- #6214 Nobody Trains an Uncensored Model From Scratch
- #6213 Ten Weirdest AI Evals Ever Published
- #6212 Woolly Sedate Snail and Other Model Failures
- #6211 Attention Is Still Winning: The Hybrid Model Census
- #6210 Who Decides If an Election Is Real?
- #6209 Watching Elections: Who Collects the Numbers, and Why
- #6208 Israel's 120-Seat Forecasting Ministry
- #6207 Two WERs, One Model: Inside ASR Eval Harnesses
- #6206 Monte Carlo: How Random Guesses Solve Impossible Problems
- #6205 Where Should Your Notebook Actually Run?
- #6204 Two Humans, One Table Each: Calling Israel's Election
- #6203 How Claude Code Sessions Message Each Other
- #6202 Israel Votes, Part 6: Asking the Decision Models
- #6201 Inside Cloudflare's Clef: A Language Model That Can't Speak
- #6200 What Israeli Election Polls Hide Behind Seat Counts
- #6199 Running AI Experiments Before an Election
- #6198 Benchmarking a TTS Model on Your Own GPU
- #6197 Talking to Your Own Voice Clone
- #6196 Fine-Tuning ASR: When Your Model Memorizes You
- #6195 Me and My Voice Clone
- #6194 Teaching a TTS Model His Own Accent
- #6193 Why Email Images Still Phone Home
- #6192 What Actually Happens When You Train a Model
- #6191 What Actually Happens During Fine-Tuning
- #6190 When Your AI API Quietly Changes Under You
- #6189 What "Real" Improvement Actually Means in AI Evals
- #6188 Every Way Linux Relieves Memory Pressure
- #6187 Can Your Laptop Borrow Your Desktop's RAM?
- #6186 A Virtual Disk Inside Your Linux Filesystem
- #6185 Why AI Labs Sell a Tier Above the Flagship
- #6184 Anthropic's Unreleased Model Two, Explained
- #6183 Private Podcast Feeds: Tokens, Buckets, and Basic Auth
- #6182 Running a Private APT Repo for Family Machines
- #6181 Shipping Private Android Apps to Your Family
- #6180 Mapping Code-Switching: 113 Pairs, No Community
- #6179 Voice Cloning: 10 Seconds to Steal Your Voice
- #6178 Why Your Voice Clone Sounds American
- #6177 How Voice Cloning Works From Six Seconds of Audio
- #6176 Do Audio Models Hear How You Said It?
- #6175 Video Models Can't Tell Time: What Motion Benchmarks Reveal
- #6174 What Your Voice Reveals: Accent, Distance, and Movement in Audio AI
- #6173 Auditing a Sycophantic Model From the Inside
- #6172 Recipes vs. Regeneration: AI Photo Editing in 2026
- #6171 AI Audio Restoration: What Actually Ships in 2026
- #6170 One Endpoint, Many Boxes: Local LLM Routing
- #6169 What It's Like to Be a Token Inside an LLM
- #6168 Hebrish: Hebrew-Aware Dictation That Actually Works
- #6167 Streaming Browser Cookies to a Headless Server
- #6166 Turning a Drifted Server Into a Rebuildable Playbook
- #6165 Watching 20 Batch Jobs at Once
- #6164 We Gave Each Host Its Own AI Agent
- #6163 Forking a Keyboard for One Voice Feature
- #6162 Why Dictated Words Rewrite Themselves
- #6161 Turning Words Into Sound: G2P, IPA, and the Hebrew Vowel Problem
- #6160 Where Do Tiny Speech Rules Live? A Ladder From Regex to LoRA
- #6159 When a Speech Model Can't Decide How to Write Dollars
- #6158 Naming the ASR Sub-Tasks Nobody Mapped