
Sign up to save your podcasts
Or


Teaser: Al Gore didn't invent the internet. But he funded it, named it, and spent thirty years trying to explain it to people who thought he was lying. This is the story of how a senator from Tennessee became the unlikely architect of the information age — and why the joke that followed him never quite died. Six acts. Starts here.
New episodes every weekday.
The final act of Pulse: Deep Dive's quantum series lands on a single arithmetic problem: if your data must stay secret for fifteen years and migration takes five, you needed quantum-resistant encryption before the conversation started. The qubit count required to break RSA-2048 collapsed from twenty million to nine hundred thousand through math alone, with no new hardware required. Mosca's inequality, the five-dollar-a-terabyte storage economics, and NIST's hard 2035 removal deadline are all running simultaneously.
00:00 — Cold open: TLS handshakes and harvest-now-decrypt-later
New episodes every weekday.
The Race Nobody Is Winning: Google claims supremacy, IBM challenges it within 48 hours, China outspends everyone by 4-to-1, and the entire field lives on margins so thin — 0.1 percentage points — that the line between breakthrough and noise is almost invisible.
New episodes every weekday.
Peter Shor published the algorithm that breaks RSA, Diffie-Hellman, and elliptic curve cryptography in 1994 — thirty years before the machine to run it exists. Intelligence agencies in at least three nations are already recording encrypted internet traffic today, storing ciphertext they cannot yet read, under a strategy the signals intelligence community calls harvest now, decrypt later. Google's December 2024 Willow chip achieved below-threshold error correction for the first time in the field's history, and the country outspending every competitor by a factor of four isn't in Silicon Valley.
01:02 — Harvest now, decrypt later strategy explained
03:45 — Feynman's 1981 counting argument: why classical computers can't simulate quantum systems
07:20 — Manin and Benioff: the independent 1980 convergence Knox calls a storm breaking over several people at once
09:10 — David Deutsch defines the universal quantum computer, 1985
10:30 — Shor's algorithm: hours versus billions of years for the same factoring problem
13:15 — One algorithm, one structural weakness, across all deployed public-key cryptography
16:30 — Decoherence: qubits that survive under a hundred microseconds
18:10 — Google's Willow chip crosses the below-threshold error correction threshold, December 2024
New episodes every weekday.
The encryption protecting your bank records, medical history, and private communications was built on a mathematical assumption that just collapsed. A 2022 algorithmic breakthrough cut the qubit threshold for breaking that encryption from twenty million to potentially ten thousand — meaning the timeline moved years forward without a single hardware improvement. Knox walks through what intelligence agencies already know, what they've already done, and why the window to act is shorter than anyone is saying.
New episodes every Saturday.
Who Gets To Have It: By 2025, the question had shifted from 'can AI work' to 'who gets to have it': DeepSeek's $6 million model triggered a $593 billion single-day crash in Nvidia's market cap, proving that cost efficiency could undermine hardware advantages; US export controls expanded to potentially cover every semiconductor tool on Earth; Anthropic became both a Pentagon contractor and a Pentagon refusenik in consecutive years; and Nvidia reached a $5 trillion valuation with a forward P/E ratio actually below its five-year average — because earnings grew faster than the most frenzied stock rally in history. The field that started with one man's thought experiment about child-machines now sits at the center of great-power competition, and no institution — not a company, not a government, not an alliance — has demonstrated the ability to control what happens next.
CHAPTERS
New episodes every weekday.
The Pathological Liar Ships Anyway: 2023 was the year AI escaped the lab and the public panicked in slow motion: ChatGPT set the all-time adoption record, Google lost $100 billion from one wrong tweet, Meta's model weights leaked and couldn't be un-leaked, GPT-4's bar exam score turned out to be a framing trick, and Geoffrey Hinton — the person most responsible for modern AI existing — resigned from Google at 75 and told the world he partly regretted his life's work. The systems were impressive, unreliable, and already everywhere.
CHAPTERS
New episodes every weekday.
Nobody Noticed: From 2012 to 2022, a single theoretical insight — how AI performance scales with data and compute — passed through five organizations on three continents like a baton no one could hold: OpenAI discovered scaling laws, the discoverers left to found Anthropic, DeepMind corrected the math, Meta applied the correction to build LLaMA. Meanwhile, all eight authors of Google's Transformer paper departed, AlphaGo's Move 37 revealed that AI could think in ways humans couldn't even parse, and GitHub Copilot — not ChatGPT — quietly became the first mass-market LLM product seventeen months before the world noticed.
CHAPTERS
New episodes every weekday.
Geoffrey Hinton left the United States in 1987 because military money in AI disturbed him — a decision that accidentally made Toronto the capital of deep learning's revival. Twenty-five years later, his student Alex Krizhevsky trained AlexNet on two consumer gaming GPUs and beat the field by a margin so large it was initially assumed to be an error: 10.9 percentage points in a competition where progress was measured in fractions. This single result ended one paradigm and started another, and it happened in a bedroom, not a lab.
CHAPTERS
New episodes every weekday.
From the publisher's feed