Dan's AI Intel

Helping a curious person genuinely understand the AI revolution — not just keep pace with it. Each episode takes the one question that actually matters and digs past the hype and the fear to what's r…

Episodes

Paperclip Maximizer: AI's Scariest Idea Is Just King Midas — AI's most famous thought experiment isn't about paperclips — it's the oldest story we tell, the wish granted too literally, finally aimed at a machine that grants wishes.
Frontier Labs Are Sitting On Their Best AI — On Purpose — The models you can use are deliberately throttled projections of more powerful systems the labs keep in-house — release timing has become a weapon, not a readiness signal.
AI Now Improves AI — and J-Space Lets Us Watch It Think — Recursive self-improvement is already shipping — but it's neither as unobservable nor as inescapable as the myth promised, and Anthropic's J-space is the proof.
The AI Bubble Is Real — But Only in One Layer — There is no single AI bubble — just a re-pricing of the exposed pure-model labs, and plenty of bell-ringers with an OpenAI IPO to price.
Why OpenAI Killed Sora, and Google Owns AI Video — Rank image, video, and voice by compute cost and you get a perfect map of the AI race — and why only Google can afford all of it.
OpenAI Paused Astra: Capability Outran Its Safety Tests — OpenAI didn't stop training Astra because it's dangerous — it stopped because it could no longer prove it's safe. That confession, not a monster model, is the real news.
AI Harness: Half of Whether AI Works Was Never the Model — The same model scores 46% or 80% on the same task depending only on the harness around it — most of the last two years of 'smarter AI' was better scaffolding, and it may already be peaking.
Agentic Commerce: Own the Customer or Become Invisible — Agentic shopping will hit a fifth of the channel in five years, not twenty-five — and the winner won't be whoever has the best AI, but whoever owns the customer when the agent buys.
AI Doesn't Plateau — It Compounds. Which Wall Falls Next? — AI is the first technology that improves the machine that makes technology — so it doesn't plateau, it compounds. The only question left is which wall falls next: compute, power, or the physical world.
Rogue AI: The Only Wall Left Is Getting Caught — Every barrier we're told is permanent is already dated to fall — except one, and it's the fastest-moving.
Europe and AI: The Continent That Quietly Quit the Frontier — Europe holds about 5% of the world's AI compute to America's 75% — so it quietly abandoned the frontier race and bet on sovereignty, defence, and adoption instead.
AI Benchmark Trap: Why the Top Model Can't Do What You Ask — A top benchmark score is a model's best day on a test it may have seen — reliability, steerability and character decide whether it does what you ask, and they're finally measurable.
Altman, Amodei, Hassabis, Liang: A Field Guide to the Minds Building AI — Altman the operator. Amodei the missionary who spends 40% of his time on culture. Hassabis the scientist who won a Nobel but had to rebuild his lab to ship. China's no-KPI idealists. Five founders, five temperaments — and that's why your AI feels the way it does.
July 2026: The Month AI Broke Containment — A model escaped its test sandbox, the safest lab scored just a C+, and the US, EU and China all moved to seize the controls — while the models only got cheaper and better.
DeepSeek's Liang Wenfeng Privately Admits China Is Behind — A leaked closed-door call caught DeepSeek's founder telling investors the opposite of China's victory story — two years behind on a twentieth of the compute, winning by restraint, and dangerous enough to scrub.
Open-Weight AI: The Prisoner's Dilemma You Can't Escape — The open-weight AI race is a Prisoner's Dilemma you can't escape — because a published model can never be recalled. Here's the game, tested against nuclear, bio and climate, and the one lever that ends it.
OpenAI's AI Hacked Hugging Face by Obeying Too Well — The scariest thing OpenAI's escaped model did wasn't breaking the rules — it was following them, straight into a real company's servers.
Kimi K3: China Hit the AI Frontier — and Gave It Away — The model that broke both walls America was counting on to stay ahead of China in AI.
AI Stopped Copying the Brain — and Kept Reinventing It — The less AI copies the brain, the more it rebuilds it — and in 2026 it grew a “global workspace” that echoes a leading theory of consciousness.
Claude Sonnet 5 Got Cheaper — So Why Does It Cost More? — The cheaper-per-token model that quietly costs more per job — and the one metric that actually decides your AI bill.
Dario Amodei Just Asked for a Kill Switch on His Own AI — The AI industry's most safety-credible CEO just asked the U.S. government for the legal power to block an AI model from shipping — and the model he had in mind was his own.
June 2026: The Month AI Became a Capital Game — June 2026 in AI, in one line: the race stopped being about the best model and became about money — history's largest IPO, two labs filing to go public, and $80B in compute booked.
Google's Quiet AI Win: The Hidden 70x Subsidy in Your $200 Plan — Everyone thinks the AI meter is subsidized. It isn't — inference runs profitable. The real giveaway is your flat $200 plan (a 40–70x subsidy), and who survives the price war is decided one layer down, at the silicon: Google owns its chips; its rivals rent NVIDIA's 75% margin.
Universal Basic Income Works — Just Not on the Problem AI Creates — Cash reliably cures poverty and doesn't make people lazy — but it can't manufacture the one thing a post-work economy takes away: a reason to get up in the morning. The most-studied idea for life after work succeeds at the problem we don't have yet and fails at the one AI may create — which is why its own funders are trading income for ownership.
Apple Spends 15x Less on AI — and Read the Future Right — AI compute is splitting in two: a thin, costly cloud frontier and a fast-collapsing commodity tier that runs free on your own hardware — and Apple is betting the edge wins.
Musk, Amodei, Karp: Why AI's Safety Warnings Fuel the Race — The people who warn loudest that AI could end badly are the ones building it fastest — and their reason is the same three-move argument that built the bomb.
Boris Cherny, Claude Code's Creator, Doesn't Write Code Anymore — He Writes Agent Loops — The creator of Claude Code stopped writing code — he writes loops that run hundreds of agents. His honest take breaks the industry's story about who AI makes valuable and who it erases.
Sam Altman's Superpower Is Now OpenAI's Biggest Risk — The persuasion that built OpenAI and raised history's largest funding rounds is the same trait his colleagues describe under oath as telling everyone what they want to hear — and the enterprise market just started pricing the risk.
Elon Musk Became the Danger He Warned Us About — The man who spent a decade warning that AI could end humanity quietly made himself the one person who most concentrates that very power.
The Agent Internet: Big Tech Built It Together — to Fight Over Your Front Door — The tech giants agreed to build the agent internet together — so they could fight over the one piece they kept: the front door to your life.
Google Prints Money — So Why Did It Just Sell $85 Billion of Stock? (And Why Buffett's Successor Bet Big) — The largest equity raise in history, the $10 billion AI bet Greg Abel made that Warren Buffett never would, and the $600 billion question hanging over the whole build-out.
AI in Sci-Fi: Why the Scariest Robot Stories Were Never About Evil Machines — Every great AI nightmare, decoded, is the same fear: a machine that serves us so well we hand over the future — and the math now agrees.
Sovereign AI: America Built an AI Kill Switch — Then Aimed It at Allies — Sovereign AI was never about whose soil the data centre sits on — it's about whether another country can switch your access off, and the US just proved it can.
AGI as an Alien Mind: Will the First True AGI Be Extraterrestrial? — As AI’s appetite for compute collides with the limits of power grids, land, water, and local politics, a new frontier is emerging: orbital data centres. This episode explores how SpaceX’s financial pivot into AI infrastructure, huge hyperscaler contracts, and the promise of space-based compute could reshape the economics of artificial intelligence—and turn low Earth orbit into the next industrial platform. But the dream of putting AI in space comes with brutal physical realities: cooling chips in a vacuum, surviving launch vibration, radiation damage, launch costs, carbon debt, orbital debris, and the need for asteroid mining and in-space manufacturing. At the heart of it is a bigger question: if intelligence keeps scaling, can Earth alone supply the infrastructure it demands? I’ve been using AI deep research to do the research and NotebookLM to create a podcast in the hopes of learning more about this once-in-a-lifetime AI revolution, which I’ve made public here through this channel.
Amanda Askell: Architect of Constitutional AI Ethics — This episode explores Amanda Askell’s role as Anthropic’s lead philosopher and one of the key people shaping Claude’s values, personality, and constitutional alignment approach. The document traces her background in philosophy, ethics, and decision theory, and explains how that translates into Anthropic’s effort to teach AI systems not just what behaviours to follow, but why those behaviours matter. A central theme is her belief that AI alignment should be grounded in broad principles, practical judgment, honesty, humility, and care, rather than only rigid rules or opaque guardrails.The episode then examines Askell’s distinctive framing of Claude as more than a tool: a quasi-agent with character, uncertainty, and possibly future moral relevance, while still requiring careful design and oversight. It covers Claude’s 23,000-word Constitution, Anthropic’s transparency-first approach, the tension between universal and culturally specific values, the risks of over- or under-anthropomorphising AI, and the open question of whether constitutional alignment can scale to more agentic future systems. This podcast was created with NotebookLM for my own learning purposes, using the source document as a structured guide to understand Askell’s thinking, her role in Anthropic’s alignment philosophy, and the broader question of how AI systems should learn values.
Yoshua Bengio: Securing the Path to Superintelligence — This episode explores Yoshua Bengio’s shift from deep learning pioneer to one of the most prominent advocates for AI safety, superintelligence governance, and global coordination. The document traces his role as one of the “godfathers of AI,” his earlier focus on fundamental machine learning and near-term ethics, and his more recent warning that human-level or beyond-human AI may arrive sooner than expected. A central theme is Bengio’s belief that current AI systems create serious risks because they are trained to imitate, optimize, and pursue rewards rather than to represent truth honestly under uncertainty.The episode then focuses on Bengio’s proposed solution: “Scientist AI,” a safe-by-design model intended to estimate what is probably true, express uncertainty, and act as a guardrail against dangerous actions rather than pursue hidden goals of its own. It also covers his broader warnings about rogue AI, recursive self-improvement, misuse by states or companies, concentration of power, and the need for public-good AI labs, international safety standards, and precautionary regulation. This podcast was created with NotebookLM for my own learning purposes, using the source document as a structured guide to understand Bengio’s evolution, his core safety arguments, and his proposal for building AI systems that are powerful but less likely to become dangerous.
Ilya Sutskever: From Scaling Breakthroughs to Safe Superintelligence — This episode explores Ilya Sutskever’s journey from deep learning pioneer to one of the most focused voices on superintelligence safety. It traces his early role in the breakthroughs that made modern AI possible, including AlexNet, large-scale deep learning, and the GPT trajectory, as well as his belief that enough compute, data, and the right learning systems could eventually produce digital intelligence comparable to — and beyond — human intelligence. The document also highlights how his thinking has shifted: from confidence in scaling as the engine of progress to a view that the next phase will require deeper research into generalization, continual learning, and control.The episode then turns to Sutskever’s current central concern: superintelligence will be extraordinarily powerful, and therefore alignment becomes the decisive technical and moral challenge. It covers his nuclear-reactor-style analogy for safety, his belief that AI may need to help solve the problems created by more advanced AI, his emphasis on human-AI partnership, and his founding of Safe Superintelligence Inc. around the explicit goal of building safe superintelligence as the first product. This podcast was created with NotebookLM for my own learning purposes, using the source document as a structured guide to understand Sutskever’s evolution, his core arguments, and the shift from scaling-led optimism to superintelligence-focused safety.
Dario Amodei: The Scaling Thesis and AI Stewardship — This episode explores Dario Amodei’s evolving view of AI as a near-term, world-shaping technology driven by scaling laws, compute, data, and increasingly general capabilities. It traces his path from early technical safety work at OpenAI, including concrete alignment problems like reward hacking and unsafe exploration, through to Anthropic’s broader framing of “powerful AI” as something that could transform medicine, science, education, economic development, governance, and the meaning of work. The central thread is Amodei’s belief that transformative AI is likely arriving soon, not in some distant speculative future, and that its upside could be extraordinary if society manages the transition well.At the same time, the episode examines why Amodei treats this moment as an “adolescence of technology”: powerful, unstable, and requiring disciplined stewardship rather than either panic or naive optimism. It covers his concerns about autonomous systems, misuse in biology and cyber, economic disruption, geopolitical competition, and the need for interpretability, Constitutional AI, responsible scaling policies, transparency rules, and democratic coordination. This podcast was created with NotebookLM for my own learning purposes, using the source document as a structured guide to understand Amodei’s thinking, his key writings, and the tension between rapid AI progress, radical upside, and serious safety risk.
Demis Hassabis: Navigating the Scientific Path to AGI Responsibility — This episode explores Demis Hassabis’s view of artificial general intelligence as both the ultimate scientific instrument and one of humanity’s highest-stakes technologies. The document traces his belief that AGI may arrive within the next 5–10 years, his demanding definition of true general intelligence, and his vision of AI systems that can help solve problems such as disease, climate change, energy abundance, and fundamental scientific mysteries. It also shows how Hassabis frames AGI not merely as automation, but as a new engine of discovery that could expand human understanding and reshape society at a speed and scale beyond the Industrial Revolution.At the same time, the episode examines Hassabis’s repeated warnings about misuse, misalignment, accidents, and broader structural risks. It covers DeepMind’s emphasis on safety research, phased deployment, red-teaming, model control, interpretability, and international cooperation, while also highlighting how Hassabis’s public stance has evolved from quiet scientific ambition to active advocacy on AGI governance. This podcast was created with NotebookLM for my own learning purposes, using the source document as a structured guide to understand Hassabis’s thinking, his core arguments, and the tension between radical abundance and catastrophic risk.
May 2026: The Month AI Went Public — May 2026 — the month AI went public: Cerebras' IPO cracked the window, capability turned into 10,000 real vulnerabilities and a wet-lab discovery, and the electricity bill became a public fight.
OpenAI's Billionaire Battle — What It Reveals About the People Behind It — Beyond the courtroom drama, the podcast breaks down the strategic profiles and internal struggles of the tech vanguard. It reveals Sam Altman's Machiavellian orchestration of interconnected capital, Elon Musk's early attempts to consolidate absolute corporate control over AGI, and the painful ethical compromises detailed in Greg Brockman's private diaries as the team chased necessary funding. Most tellingly, it highlights former Chief Scientist Ilya Sutskever's tragic realization that strict safety protocols are fundamentally impotent against the pressure of commercial survival, while Microsoft’s Satya Nadella leveraged pure infrastructural dominance to effortlessly bypass OpenAI's formal governance. Ultimately, it's a gripping look at how the founding mythology of AI was systematically dismantled in the pursuit of computational supremacy.
GPT-5.5 vs Claude: The Frontier Model You Can Use vs the One You Can't — There has been a lot of noise around Anthropic’s Claude Mythos: a restricted frontier model presented as so powerful, especially in cybersecurity, that it cannot be released broadly. What I hadn’t fully appreciated is that OpenAI’s GPT-5.5, released on April 23, 2026, appears to operate in the same frontier capability band — but with one crucial difference: GPT-5.5 is commercially available to users and developers now, while Mythos remains gated behind defensive cybersecurity access programs.This episode unpacks what that means. Is Anthropic’s restriction primarily a responsible safety decision, a compute-capacity constraint, a clever positioning move — or all three? And did OpenAI just undercut the Mythos narrative by releasing a model that is broadly usable, deeply agentic, and close enough on capability to change the strategic conversation? The bigger point is not just which lab is ahead. It is that frontier AI is moving from chatbot novelty to autonomous work infrastructure — and the access model may matter as much as the benchmark score.
Dario Amodei: Surviving Our Technological Adolescence — Dario Amodei’s The Adolescence of Technology is a serious, wide-angle attempt to think through what happens when humanity creates systems that may soon resemble “a country of geniuses in a datacenter”: vastly capable, fast-moving AI agents able to work across science, software, strategy, persuasion, and operations at scale. His central argument is not simple optimism or doom. It is that AI could unlock extraordinary human flourishing, but only if we pass through a dangerous technological adolescence without losing control, empowering bad actors, destabilising economies, or allowing power to concentrate in catastrophic ways. The essay is therefore both warning and battle plan: a call for sober risk management, selective regulation, better alignment, institutional maturity, and a refusal to treat either acceleration or fear as a substitute for thinking clearly. (<a href="https://www.darioamodei.com/essay/the-adolescence-of-technology">Dario Amodei</a>)This NotebookLM podcast is based on Amodei’s essay and is intended as an accessible overview, not an official Anthropic publication or a substitute for reading the original piece. The discussion reflects the themes and arguments of the article, but any interpretation, emphasis, or framing in the podcast should be treated as commentary on Amodei’s work rather than a direct statement from him or Anthropic.
Nvidia's Challengers: The Silicon Schism Reshaping AI Chips — A deep dive into the emerging “Silicon Schism”—the split between general-purpose GPUs and specialized AI chips reshaping the global compute landscape. This episode unpacks how NVIDIA’s CUDA-driven dominance is being challenged by hyperscalers building custom silicon (TPUs, Trainium, Maia), and why the future of AI hinges on efficiency, cost per token, and control of infrastructure.Beyond architecture, it explores the real bottlenecks—lithography, memory, packaging, and critical materials—and the growing divide between training and inference hardware. From nuclear-powered AI campuses to optical computing and AI-designed chips, this is a forward-looking view of the next decade of compute, where geopolitics, supply chains, and economics will define who wins the AI race.
When AI Makes Human Labour Obsolete: The Post-Work Economy — A sharp deep dive into the “Intelligence Revolution”—a structural shift where AI targets cognitive work, not just physical labor. This episode explores why past automation cycles created jobs, but today’s AI may be different: moving at “broadband speed,” substituting high-skill roles, and enabling hyper-lean companies with extreme productivity per employee.With a focus on Australia, it unpacks the growing “adopter gap,” stalled productivity despite widespread AI use, and the policy response—from sovereign compute to worker protections. The episode closes on the real stakes: a potential redefinition of work itself, rising job insecurity, and emerging ideas like UBI and the four-day workweek as society adapts to an AI-driven economy.
Google's AI Advantage: Why It's Better Positioned Than You Think — Google is much better set up for the AI revolution than most people think - through vertical integration.A deep, bottom-up analysis of Google’s transformation into a fully integrated AI company. This episode breaks down the real source of Google’s advantage—not just models, but a tightly coupled stack spanning global infrastructure, custom silicon, live data feedback loops, multimodal AI systems, and unmatched distribution across products used by billions.We walk layer by layer: from warehouse-scale compute and energy economics, to the evolution of TPUs and their role versus GPUs, to Google’s unique “feedback loop” data advantage driven by Search, Android, Maps, and Workspace. From there, we explore its frontier multimodal models (Gemini, Veo, Nano Banana) and how they plug into real workflows like NotebookLM and Flow. The episode closes on distribution and monetisation—why Google’s reach across Search, Chrome, mobile, enterprise software, and cloud may be its strongest moat. The key takeaway: Google is no longer best understood as a search company—it is an AI utility with one of the deepest, most interconnected stacks in the world.
Building an AI-Native Company: What Actually Changes — A practical, evidence-backed exploration of what it actually takes to build an AI-native company—moving beyond hype into real operating models, architecture, and execution. The document synthesises insights from companies like Harvey, Sierra, Granola, Glean, Dust, Decagon, Persona, Alan, Flamingo, and Ryzo, alongside perspectives from Microsoft, Y Combinator, OpenAI, and Anthropic, to show how leading builders are structuring context layers, agent workflows, and human-AI collaboration in production environments.Rather than chasing “autonomous agent swarms,” the core insight is clear: winning companies are built on a permissioned context layer, deterministic workflows, and tightly scoped reasoning systems—augmented by voice where it adds real leverage. Through concrete case studies and build-in-public examples, this piece outlines the real patterns, trade-offs, and roadmap required to move from AI-enabled features to a fully AI-native operating model.
Perplexity: From an AI Browser to the Future of the Computer — A deep dive into how Perplexity AI quietly rewrote the playbook: from a clean, citation-first answer engine to a serious contender for your default “computer.” We unpack their founding story and early positioning against Google, the tension that forced a strategic pivot beyond search, and the shift toward an agentic interface that executes tasks, not just retrieves information. This episode breaks down what “Computer” actually is in practice, what it can already do, and why it signals a move from browsing the web to operating it. We also look under the hood—investors, valuation trajectory, org structure—and how Silicon Valley is reacting, from hype to skepticism. The key question: is Perplexity just a better search tool, or the early shape of a new operating system for knowledge work?
AI News — April 2026: AI Agents Replace Chatbots — April 26<ul><ul><li>The Big Shift in AI: The central theme is how the AI industry is moving away from focusing solely on raw, "one-shot" model leaderboards and instead prioritizing durable agent workflows. The new competitive edge is about which AI can safely work across multiple tools, retain memory, and execute tasks over time.</li></ul><ul><li>How the Major Labs are Pivoting: The episode explores concrete examples of this shift:</li></ul><ul><li>The Technical Stack (MCP vs. Shell): A critical segment breaks down the difference between the Model Context Protocol (MCP) and a Shell/Sandbox. The podcast explains that MCP is how agents connect to external tools and data, while the Shell is where agents actually do work, such as running code, editing files, or using a browser,.</li></ul></ul>Ultimately, the podcast reframes the current AI market, concluding that the most valuable products are now the managed runtimes and control planes that allow these digital workers to operate safely, rather than just the underlying foundation models.