#6074: The Worst Documentaries of 2026 (So Far)

Awards for the year's worst documentaries: recycled premises, infomercials, a two-week Instadoc, Mr. T's verdict on his own film, and Melania's 90-point critic-audience gap.

journalism-and-mediafilm-and-televisionnutrition-and-diet

#6073: Israel Votes, Part 4: The Forecast

Three forecasting methods, seven synthetic voters and one prediction for Israel's 27 October election: Yashar 21, Likud 20, and nobody reaches 61.

electionsisraelceasefires-and-peace-deals

#6072: China Beyond Beijing and Shanghai: A Regional Guide

Most Westerners know two Chinese cities. Here's the rest of the map — regions, megacities you've never heard of, and which ones you can actually re...

podcastingtransformer-architecture

#6071: Outside the English Bubble: Where to Find Dubbed Asian Documentaries

There's no Netflix for dubbed world documentaries. But AI dubbing on Bilibili and YouTube is about to unlock the part of a country that talks to itself.

ai-alignment

#6070: The Papers That Moved AI: From AlexNet to This Year's Fights

The classics all said 'learn it from data and scale it up'. This year's most-discussed AI papers are auditing what scaling took for granted.

transformer-architecturetraining-datacomputer-vision

#6069: The Whole Machine: How an Episode Gets Made, Send to Publish

The complete tour of the pipeline that makes the show, stage by stage, from the send form to publishing, including the new eight-turn batched voice decode.

workflow-automationmulti-agent-systemslocal-ai

#6068: Chinese AI Models: Refusing vs. Reframing

New studies show Chinese AI models have shifted from refusing sensitive questions to answering them — in the Party's voice.

ai-biasai-alignmentdigital-privacy

#6067: Israel's Freedom of Information Law, Explained

Israel's FOI law promises documents in 30 days. So why do ministries lose 88% of the transparency cases brought against them?

bureaucracylegislationisrael

#6066: Israel Votes, Part 2: What the Parties Stand For

Most Israeli parties haven't published a platform in years. Here's what they actually say they'll do — and where they stay silent.

israelelectionsharedi-community
Also available in:עברית

#6065: Israel Votes, Part 3: A Renter's Voting Recommendation

A Jerusalem renter wants to know which party actually helps him. We scored Israel's parties against his views on rent, prices, and daily life.

electionsisraelcost-of-living

#6064: Israel Votes, Part 1: Who Is Actually Running

Thirty-eight lists, a first on-time election since 1988, and a ballot full of parties that didn't exist eighteen months ago. Here's the map.

israelelectionsceasefires-and-peace-deals
Also available in:עברית

#6063: Ten True-Story Films for Daniel: Platforms, Newsrooms, and a Man Living in a Toy Shop

Ten true-story films from 2025 and 2026 picked for one listener, from Facebook whistleblowers to a man who secretly lived inside a Toys R Us.

#6062: Who Actually Shapes Chinese AI Thinking?

Beyond the lab CEOs: the researchers and ideas defining how China builds — and argues about — artificial intelligence.

chinatraining-dataai-alignment
Also available in:中文

#6061: MARPE, Tirzepatide, and AI on CBCT: Sleep Apnea's New Tools

MARPE's bone share, the first real MARPE sleep apnea trial, and what tirzepatide did to oral appliances — plus AI reading sutures on CBCT.

medieval-world

#6060: When the Government Sues the Thing Itself

Barrels of soda, a book, clacker balls, and a family's savings — all defendants in American courts. How in rem jurisdiction works.

courts-and-judiciaryfree-speechbureaucracy

#6059: Ten Documentaries for Daniel: AI, Power, and Who Holds the Levers

A hand-picked slate of 2026 documentaries on AI anxiety, platform power, intelligence, and media dynasties — plus where to actually watch them.

ai-alignmentai-evaluationgenerative-media

#6058: The Most Lamentable Comedy of the Overlong Episode

A language model writes a Shakespearean drama for three voices — and Chatterbox TTS has to say "forsooth" without falling over.

podcastinggenerative-medialanguage-learning

#6057: Why Your Recorded Voice Sounds Like a Stranger

Bone conduction, brain prediction, and identity — why hearing your own voice on tape makes you cringe, and what changes when AI can speak for you.

sensory-processingrelationshipspsychiatric-medication

#6056: Where Does Speaker Structure Belong in AI Podcast Scripts?

Label-and-parse pipelines break in ways that look like model errors but aren't. Plus what the research actually says about the format tax.

prompt-engineeringai-evaluationhallucinations

#6055: How to Fine-Tune ASR on Noisy Audio

A small noise library mixed into clean speech beats collecting hours of noisy recordings. Here's the ratio that actually works.

speech-recognitiontraining-datafine-tuning

#6054: Should Citizens Get an SLA on Government Data?

A decade of open data portals moved the share of fully open datasets from 7% to... 7%. What would a real publishing pipeline look like?

open-sourcebureaucracylegislation

#6053: Köppen's 30 Boxes vs. a Changing Planet

Köppen squeezes the planet into 30 boxes — and the hottest places can't change class, even as they heat. What the map hides.

climate-changetaxonomy-and-classificationweather-and-atmosphere

#6052: What's Actually in the Hugging Face ASR Catalog?

Four real architectural bets sit underneath thousands of Hugging Face ASR repos. Here's how to tell the bets from the repackagings.

fine-tuninglocal-aitransformer-architecture

#6051: How to Find the AI Benchmark You Actually Need

Thousands of benchmarks, no canonical index. How Benchmark Radar and the Lighteval Tasks Explorer help — and where they fall short.

ai-evaluationlocal-aiopen-source

#6050: Codex vs Claude Code: What the Numbers Actually Count

Codex has more users, Claude Code has more revenue — both can be true. The difference is what each number is counting.

ai-coding-assistantsai-evaluationcontext-and-memory

#6049: Claude Code by the Numbers: What's Measurable, What's Hype

Anthropic won't publish a user count. Here's what revenue, surveys, and 400K sessions actually tell us about Claude Code's rise.

ai-coding-assistantsai-evaluationtraining-data