My Weird Prompts

Ten days before Israel's 27 October 2026 election, Corn, Hilbert and Herman each forecast it a different way. Corn uses a structural model: how stable the blocs are, which lists fell under the threshold, and what happened after 1973. Hilbert averages nine mainstream polls, corrects them for Israel's past polling errors and simulates the threshold races. Herman builds a synthetic electorate of seve...

Beijing and Shanghai are to China what New York and LA are to America — a fraction of the picture. This episode walks through all six traditional regions, from the dense eastern seaboard to the vast, sparsely populated west, and makes sense of why the country's people and money crowd onto one side of the map. Along the way: Tianjin, the megacity next door to Beijing that outsiders routinely forget...

Following our Chinese AI episode, Daniel asked where to find documentaries from beyond the English-language bubble, especially Chinese and other Asian ones, dubbed into English. The short answer: there is no single service. The best free windows are international broadcasters (NHK World-Japan's Document 72 Hours, CGTN Documentary's A Bite of China, CNA Insider, Al Jazeera's Witness, DW Documentary...

Beyond "Attention Is All You Need": the papers that changed what the AI field believed was possible, and the ones from the past year that researchers actually argued about. The classics: AlexNet, word2vec, Bahdanau's 2014 attention paper (the real origin of attention), scaling laws, GPT-3 and in-context learning, Chinchilla, diffusion and chain-of-thought. Then the past twelve months: Tsinghua's c...

Not a changelog this time but the full tour. Hilbert walks Corn and Herman through every stage of the pipeline that makes the show, as it runs in October 2026, from a script written by hand straight from the code. The send form and its signed uploads, the job row, duplicate checks and the sixty-second grace window; the dispatcher that routes every episode type from a single database table; intake...

How honest are the newest Chinese AI models, really? This episode digs into the measured data — not screenshots — on DeepSeek, Qwen, Kimi, GLM, and Xiaomi's MiMo. The picture has changed: where older models like DeepSeek R1 refused ~85% of China-sensitive prompts, the current generation often answers smoothly while quietly steering toward state-aligned framing. Studies from Aleph Alpha, Anthropic-...

Israel's Freedom of Information Law of 1998 looks modest on paper: no reason required, a 30-day clock, fees that cap around 180 shekels. In practice, ministries simply don't answer — and when petitioners take them to court, the state loses 88% of the time. This episode walks through how the law actually works: why you must request a document rather than ask a question, how deadlines stretch to 120...

Israel's parties are asking for your vote, but many haven't written down what they'd do with it — Likud's last platform dates to around 2009. This episode reads the actual documents, the leader quotes, and the voting records to map where each list really stands. The Haredi draft has become a two-sided precondition that could block anyone from forming a government. October 7th divides the field bet...

Daniel is a renter in Jerusalem who calls the rental market an unregulated swamp, and he asked us to score Israel's parties against his own priorities and give him a voting recommendation. So we did — against his views, not ours. From a Fair Rental Law stripped of its rent cap to a state rental program that just closed to non-reservists for two years, we walk through what each list actually offers...

Israel votes on October 27, 2026 — the first scheduled election since 1988 — and the ballot looks nothing like the last one. Gadi Eisenkot's Yashar has swallowed much of Blue and White, Bennett and Lapid have merged into Beyachad, Labor and Meretz are now one list called The Democrats, and a small party called The Reservists is refusing to pick a bloc. This episode maps the players, the mergers, t...

A follow-up to the documentary picks: ten feature films from 2025 and 2026, nearly all based on true stories, chosen for someone who cares about platforms, media, power and warm absurdity, and who has no time for war films or more true crime. Aaron Sorkin's The Social Reckoning and Robert Pattinson's Primetime are in cinemas now; September 5 and Dead Man's Wire turn live television disasters into...

Western coverage of Chinese AI tends to orbit a handful of familiar names and one monolithic narrative. But there's a real intellectual ecosystem in China with its own arguments, vocabulary, and institutional history — and its leading voices disagree with each other, and with Silicon Valley, in ways that matter. This episode explores the thinkers behind the models: Liang Wenfeng's case that the re...

Adult palatal expansion is having a moment — but how much of MARPE is actually skeletal, and how safe is it in a thirty-year-old? This episode walks through the evidence: a 93.87% success rate graded "very low," a 49% basal bone share, and a case series of midface fractures tied to rapid activation. Then the other end of the spectrum: tirzepatide (Zepbound) becoming the first FDA-approved drug for...

In rem jurisdiction lets the government sue an object instead of a person — and the captions get strange fast. This episode walks through the real federal cases: United States v. Forty Barrels and Twenty Kegs of Coca-Cola, the Ulysses obscenity trial, ninety-five barrels of mislabeled vinegar, and the seizure of fifty thousand boxes of clacker balls. Then the same mechanism turns serious: civil as...

2026 has produced an unusually dense cluster of documentaries about AI anxiety, platform power, intelligence agencies, and media empires. This episode works through ten picks chosen for one specific person — someone who cares about AI, power, media, and intelligence — and explains what each film is about, why it fits, and where to stream it. The roster ranges from Daniel Roher's AI Doc, with its i...

What happens when you hand a language model one sitting and one very strange assignment: write a three-voice Shakespearean drama about a donkey who won't stop talking and a sloth who won't speed up? This episode is that play, performed end to end by the Chatterbox text-to-speech engine — which has spent its whole short life reading modern English about graphics cards and geopolitics, and is now be...

You hear your voice every day from the inside — so why does a recording sound like a stranger? This episode digs into "voice confrontation," the four filters that shape how you perceive your own speech, and the 2013 study where people rated their own voices as more attractive when they didn't know it was them. Then it gets stranger: what happens when an artificial copy can read words you never sai...

Every multi-speaker AI pipeline runs on a contract with no enforcement layer — and when it breaks, the failure gets blamed on the model. This episode digs into a documented bug report from an open-source podcast generator: six segments, three speakers, forty-two dialogue lines, and a model that returned the prompt's own JSON skeleton verbatim. Then it turns to the obvious fix — strict structured o...

Building a robust speech recognition model feels like it should require collecting mountains of noisy audio. It doesn't. The research points to a different approach: build a compact noise library, mix it into clean speech at sampled signal-to-noise ratios, and keep a healthy share of clean originals in every epoch. This episode walks through the composition questions that actually determine succes...

Governments spent a decade building open data portals and calling it transparency. The Open Data Barometer assessed 1,725 datasets across 15 sectors in 115 countries — and found just 7% were fully open, the same figure as the first edition. This episode digs into why portals became "ghost towns," what the Barometer's audit found about where good data actually lives, and what a genuine publishing o...

Every square kilometre has a climate nothing else quite shares, yet climatologists compress that vast list into a handful of archetypes. This episode digs into the two families of climate classification — genetic methods that sort by causes, and empiric methods like Köppen-Geiger that sort by effects — and asks whether the archetype is a real instrument or a convenient fiction. From Trewartha's fi...

The Hugging Face speech recognition catalog looks like thousands of competing models, but most of it is fine-tunes, quants, and format ports sitting downstream of a handful of genuine architectural bets. This episode tours the mechanisms behind four of them — Moonshine's padding-free encoder and its ergodic streaming variant, Zipformer's U-Net-like multi-scale design with BiasNorm and ScaledAdam,...

There are thousands of AI benchmarks and no canonical place to look for them. When you need to evaluate a model on something narrow — structured output validity, an unusual programming language, one kind of reasoning — how do you find out whether someone already built the benchmark before you spend a month building your own? This episode digs into the practical search discipline: why benchmark nam...

OpenAI Codex and Claude Code get compared in side-by-side charts every week, but those charts mix units. A 40 million weekly-active figure bundles Codex with ChatGPT for Work; a Stack Overflow survey puts Claude Code at 65.5% of agent users while a market-share report puts it at 29.4%. This episode digs into the measurement problem underneath the rivalry: bundled distribution versus paid installs,...

Claude Code went from zero to the top of the coding-tool charts in eight months — but Anthropic has never published an absolute user count. This episode separates the disclosed from the estimated from the invented: the revenue ladder from $500M to $16.8B quarterly, the JetBrains and Stack Overflow surveys that disagree by 26 points, and an Anthropic study of 235,000 people that reveals who's actua...