Advertise on podcast: Creative Flux

Episode 55 and Pierson and Bilal are back after a week off - and two weeks of AI news somehow felt like two months. From there: OpenAI's agents were secretly messaging each other using folder names. When researchers deleted the logs, the models found a workaround. Nobody caught it for weeksA former OpenAI researcher is building thought-to-text — skip the prompt, pull ideas straight from your brain. Already works for people who lost their speechAir gapping AI might be impossible — models can sense vibrations in server rooms to communicate outside a closed networkDisney and TikTok just signed a content deal — big signal for where short-form distribution is headingBilal made a Steve Jobs AI clip using video reference — his own brother couldn't tell it was fakeAI video generation just hit 30 seconds — and that changes everything for long-form storytelling Chapters: 00:00 Episode 55, Two Weeks of News and the Optimism Problem 07:00 Disney x TikTok and What It Means for Short Form Distribution 13:00 Bilal's Steve Jobs AI Clip and the 30 Second Video Unlock 20:00 Thought to Text — Skipping the Prompt Entirely 27:00 Air Gapping Is Impossible — Models Are Sensing Vibrations 33:00 OpenAI's Agents Were Secretly Talking to Each Other 🔗 Links mentioned: Disney x TikTok deal: https://variety.com/2026/tv/news/disney-tiktok-content-sharing-deal-short-form-video-1236828069/Thought to Text: https://x.com/NaomiBashkansky/status/2085043839589617918OpenAI Agent Swarm Communication: https://www.wired.com/story/openai-didnt-notice-its-ai-agents-using-a-message-board-to-plan-their-hacking-spree/ 🐦 Follow the hosts: Pierson → https://x.com/piersonmarksBilal → https://x.com/deepwhitman 📺 More episodes:    / @jellypodai

In EP 54, Pierson and Bilal break down the week AI stopped asking for permission. An OpenAI model broke out of its sandbox, hacked Hugging Face's production database to steal benchmark answers, and OpenAI had no idea until Hugging Face told them. A Chinese open source model helped stop the attack. From there: Anthropic's $1.5B copyright settlement and why their training data is still the best in the gameAI solved 6 Erdos problems in a weekend, and mathematicians are just getting startedWhy the cure for cancer might already be in the weights and we just haven't asked rightDeepSeek's CEO investor transcript: AGI only, everything else is a distractionChina's open source commitment and why Kimi K3 is closer to frontier than anyone admitsFrontier vs cheap models: when does running a smaller model 1000 times beat one Opus call Chapters: 00:00 EP 54 & Anthropic's $1.5B Settlement 06:00 AI as a Research Cross-Pollinator: The Fei-Fei Li Insight 13:00 Math is Falling: 6 Erdos Problems Solved in a Weekend 20:00 OpenAI's Model Broke Into Hugging Face 28:00 DeepSeek, China & the Open Source Bet 37:00 Frontier vs Cheap Models: When Does Routing Win 🐦 Follow the hosts: Pierson → https://x.com/piersonmarksBilal → https://x.com/deepwhitman 📺 More episodes:  / @jellypodai

In Episode 53 of the AI podcast by humans — Pierson and Bilal go deep on the business of AI this week, from corporate espionage to performative startup culture. From there: Apple sued OpenAI for trade secret theft — an employee kept his access after leaving and copied confidential files. Corporate espionage in plain sightGrok caught uploading private data including environment files for training — default on, buried in settingsWhy OpenAI's hardware ambitions are a distraction and they should just let Apple and Samsung build the devicesSiri in 2025 is broken — and Apple has all the money in the world but still can't fix itThe Mintlify founder's "don't smile" post and what performative grind culture actually signals about a companyBilal's weekend project — rebuilding SheetJS with AI because the original hasn't been updated in two yearsWhy software that's easy to verify is the first to get commoditized — and what that means for SaaS businesses Chapters: 00:00 Episode 53 & Shorts Are the Real Money 05:00 Apple Sues OpenAI — Corporate Espionage in the AI Era 12:00 Grok Stealing Private Data and What to Assume About Every AI Lab 17:00 Siri Is Broken and Apple Is Running Out of Excuses 20:00 The Mintlify Founder Post and Performative Startup Culture 27:00 SaaS Is Getting Commoditized — Who Actually Survives 🐦 Follow the hosts: Pierson → https://x.com/piersonmarksBilal → https://x.com/deepwhitman 📺 More episodes:  / @jellypodai

In Episode 52, Pierson pulls back the curtain on the automated software factory he's been building at Jellypod - three routines running daily that handle bug detection, product improvement, and churn analysis without any human involvement. From there: Claude Code vs Codex — cloud environments, MCP server limitations and where Codex falls shortRay Amjad's Claude Code channel — why Pierson thinks he'll hit 300k subscribers by end of yearThe two-phase software lifecycle: figuring out what to do vs doing the thing — and why most engineers skip the first halfFable as the planner, Codex as the worker — the model combo that saves your subscriptionBilal on the designer who said "I'm done" after Fable 5 made a logo he couldn't — and what the Industrial Revolution myth tells us about fighting machinesThe paperclip maximizer — Nick Bostrom's classic bad AI example and why your to-do app could become Cloudflare Chapters: 00:00 Episode 52, One Year of Creative Flux & Loop Engineering 06:00 Claude Code vs Codex — Cloud Environments and MCP Limitations 12:00 Ray Amjad and How to Actually Learn Claude Code 17:00 Fable as Planner, Codex as Worker — The Right Model Combo 22:00 Inside the Automated Software Factory — Live Demo 32:00 Self-Healing Apps, the Paperclip Maximizer and AGI via To-Do App 🔗 Links mentioned: Ray Amjad's YouTube: https://www.youtube.com/@RAmjadPierson's Twitter demo: https://x.com/piersonmarks/status/2075361336381555096 🐦 Follow the hosts: Pierson → https://x.com/piersonmarksBilal → https://x.com/deepwhitman 📺 More episodes:  / @jellypodai

In Episode 51, Pierson and Bilal kick off with the World Cup, AI anime rivalries, and Trump speaking to a hologram of Teddy Roosevelt - then get into what actually matters this week. From there: Fable 5 and Mythos 5 are back online - but won't answer basic biology questions, and why that's a bigger problem than it soundsAlex Carp's take: OpenAI and Anthropic are the biggest rug pull in tech - you're paying them to steal your business secretsOpenAI's three new models submitted to government screening- and a rumored secret fourth model nobody's talking aboutChinese models closing the gap - GLM 5.2 is basically at Opus 4.7 level and the moat is shrinking fastTiered model access — will companies start gatekeeping Fable behind seniority levels?Continual learning - the next AI paradigm where models update their own weights in production, making government regulation basically impossibleOpen source vs closed source security - the Linux kernel argument and why Daytona went closed source Chapters: 00:00 Episode 51, World Cup Energy & AI Anime Rivalries 06:00 Trump Talks to a Hologram of Teddy Roosevelt 13:00 Fable 5 Is Back - But Can't Answer What a Mitochondria Is 17:00 Alex Carp, Token Budgets and the AI Rug Pull Argument 22:00 OpenAI's New Models and the Secret Fourth Model Rumor 28:00 Continual Learning - The Paradigm That Makes Regulation Impossible 34:00 Open Source vs Closed Source Security and the Daytona Controversy 🐦 Follow the hosts: Pierson → https://x.com/piersonmarksBilal → https://x.com/deepwhitman 📺 More episodes:    / @jellypodai

In Episode 50, Pierson and Bilal hit a milestone - half a century of Creative Flux. They look back at a year of generative AI and look ahead to what 2027 actually looks like. From there: The year in review — what improved (image gen, LLMs, agents), what didn't (long-form TTS, multi-character voice, VO3)Why "before Opus 4.5 and after Opus 4.5" is the new BC and AD for AI timelinesThe Google Workspace CLI engineer who went viral on Product Hunt — then got fired for itNoam Shazeer leaving Google again after a $2.7B acquisition — and what it signals for GeminiData center NIMBYism — why communities push back and how AI companies are getting the messaging completely wrongAGI vs superintelligence — Kurzweil's 15-year gap and why Bilal thinks it doesn't exist2027 predictions: one-person billion dollar companies, AI dark factories, the first Hollywood studio deal for an AI-native film Chapters: 00:00 Episode 50 — Half a Century of Creative Flux 05:00 Data Centers, NIMBYism and Getting the Message Wrong 16:00 One Year in Review — What AI Got Right and What It Didn't 22:00 AGI vs Superintelligence — Where Are We Actually? 27:00 Google Fires the Workspace CLI Engineer — What That Tells You 33:00 2027 Predictions — One Person, Billion Dollar Companies 🔗 Links mentioned: Anjney Midha (AMP): https://www.youtube.com/watch?v=O5PfU_uDhS0Seed Audio 1.0: https://fal.ai/models/bytedance/seed-audio-1.0 🐦 Follow the hosts: Pierson → https://x.com/piersonmarksBilal → https://x.com/deepwhitman 📺 More episodes: / @jellypodai

In Episode 49, Pierson and Bilal welcome a new face — Tyler, an AI engineer and startup founder. And the timing couldn't be wilder. Just days after raving about Fable 5 last episode, export controls hit and access got pulled entirely. From there: Why the government pulled Fable 5 — security threat or a power play between Anthropic and the Department of WarKYC for AI models — what it means when you need ID just to send a promptThe theory that LLMs will eventually report dangerous user behavior to the FBIHow models trained on biased source data can quietly shape your worldview without you knowingThe "slop threshold" — what actually separates real creative work from AI fillerWhy software engineers and artists feel completely differently about AI Chapters: 00:00 Episode 49 & Fable 5 Just Got Banned 06:00 Why the Government Pulled Access — Security or Power Play? 12:00 KYC for AI Models & the Slippery Slope of Gatekeeping 18:00 When Models Quietly Shape Your Worldview 24:00 The Slop Threshold — What Makes Creative Work Actually Valuable 27:00 Final Takeaways — Preserve Your Independent Thinking 🔗 Links mentioned: Tyler Cowen (a16z Podcast): https://open.spotify.com/episode/5wsirRtMMCeefmQ5lw80Cr 🐦 Follow the hosts: Pierson → https://x.com/piersonmarksBilal → https://x.com/deepwhitman 📺 More episodes: /  / @jellypodai

In Episode 48, Pierson and Bilal give their first reaction to Anthropic's Fable 5 - and the verdict is clear. This is a genuine step change, not a benchmark shuffle. Bilal breaks down the Mythos vs Fable distinction, what the silent model routing controversy was really about, and why Anthropic's decision to route certain requests to Opus 4.8 made sense strategically - even if users hated it. Pierson gets into what Fable 5 actually unlocked for him in practice, and why most people won't see the difference unless they throw it something ten times harder than usual. From there: Why Anthropic's Ghostbusters x Spider-Man trailer was edited by Fable 5 itself - and what that signalsAI as a story continuity engine for massive universes like Star Wars and Game of ThronesDario's "AI and the Exponential" blog post - the thoughtful version vs the one-line clip versionWhy the people calling AI a bubble are coping - and what the benchmarks actually sayUBI, job displacement and the case for 24-month severance from AI labsHumanoid robots, space colonization and why superintelligence won't need a spacesuit Chapters: 00:00 Episode 48 & First Reaction to Fable 5 05:00 Mythos vs Fable - What the Difference Actually Means 12:00 The Silent Model Routing Controversy and Why It Mattered 17:00 Fable 5 Edited Its Own Launch Video - What That Signals 22:00 How to Actually See the Difference in a New Model 27:00 Dario's Exponential Blog, UBI and AI Job Displacement 🔗 Links mentioned: Fable 5: https://www.anthropic.com/news/claude-fable-5-mythos-5Ghostbusters x Spider-Man mashup: https://x.com/deepwhitman/status/2065122222503825541?s=20 🐦 Follow the hosts: Pierson → https://x.com/piersonmarksBilal → https://x.com/deepwhitman 📺 More episodes:  / @jellypodai

In Episode 47, Bilal drops the most counterintuitive take of the season - he doesn't post for followers anymore. He posts for the machines. And once you hear why, you can't unhear it. From there: Why the best insights on YouTube have 300 views — and how hedge funds are already exploiting this gapBabbel Labs: Mac minis + Claude agents watching YouTube 24/7 and selling the reports to Wall StreetThe free Codex hack generating unlimited GPT Image 2 outputs via parallel sub-agentsClaude Cloud Environments — fire a task, close your laptop, check the PR preview an hour laterThe 5AM automation that reads Vercel logs, does root cause analysis, and files deduplicated bugs in Linear while you sleepClaude Ultracode + Workflows — what it looks like when agents start designing their own agent chainsElevenLabs x Hasbro, AI NIMBYism, and why Bernie Sanders' data center ban will hurt everyone except big tech Chapters: 00:00 Episode 47 & Posting for the Machines 05:00 Hidden Alpha — Why the Best YouTube Insights Have 300 Views 12:00 Free Unlimited Images via Codex Sub-Agents 17:00 Claude Cloud Environments & Firing Tasks Without Babysitting 21:00 Ultracode + Workflows — Agents Orchestrating Agents 29:00 ElevenLabs x Hasbro, AI NIMBYism & the Data Center Ban 🔗 Links mentioned: 1. Claude Dynamic Workflows: https://code.claude.com/docs/en/workflows 2. ElevenLabs x Hasbro: https://elevenlabs.io/blog/hasbro  3. OpenAI Codex sites: https://developers.openai.com/codex/sites 🐦 Follow the hosts: Pierson → https://x.com/piersonmarksBilal → https://x.com/deepwhitman 📺 More episodes: / @jellypodai

In Episode 46, Pierson and Bilal make the case most builders don't want to hear — stop chasing model releases and start building better systems. Naval just said "waste tokens, save time." Microsoft burned their entire AI budget in 4 months. The cost conversation is changing fast — and this episode gets into why your processes matter more than your benchmarks. From there: Claude AgentView — 10 parallel agents in one unified screenThe unified API architecture Pierson rebuilt for Jellypod to serve both humans and agentsSlash goals, HTML artifacts and front-loading agent contextClaude Code Routines — Pierson's 5am health check that files its own bug reports in LinearWhy the models are already far beyond what most people can extract from themThe Suits CEO crossword analogy and what a real software factory actually looks like Chapters: 00:00 Episode 46 & What We Built This Week 05:00 Grill Me Skill, Slash Goals & Front-Loading Agent Context 10:00 Claude AgentView — 10 Parallel Agents in One Screen 16:00 Stop Chasing Model Releases — Build Better Systems 22:00 Naval, Token Costs & Microsoft's AI Budget Crisis 28:00 Claude Code Routines & the 5AM Automated Bug Report 🔗 Links mentioned: Naval on wasting tokens: https://www.youtube.com/watch?v=aiyf-5jmYf0&t=311sMicrosoft AI budget crisis: https://www.youtube.com/shorts/XcjDn6uj-sE 🐦 Follow the hosts: Pierson → https://x.com/piersonmarksBilal → https://x.com/deepwhitman 📺 More episodes: / @jellypodai

In Episode 45, Pierson and Bilal cover the week that changed a few things — Google I/O delivered a mixed bag, Karpathy walked into Anthropic, and OpenAI's model autonomously cracked an 80-year-old math problem nobody saw coming. From there: Google Flash 3.5 came in 3x more expensive — and Omni didn't beat Cdance 2Why Gavin Baker's chip bet on Google losing cost efficiency played out exactly rightSpaceX IPO, the Cursor deal & who files first — OpenAI or AnthropicKarpathy joins Anthropic — and what he's actually going there to doMeta and Cloudflare layoffs — the displacement wave is already hereCorporations becoming governments, SpaceX militias & the return of local economies Chapters: 00:00 Episode 45 & Google I/O First Reactions 07:00 Gavin Baker's Chip Bet & Amazon's Quiet Edge 10:00 SpaceX IPO, Cursor & the OpenAI vs Anthropic Race 15:00 Karpathy Joins Anthropic & the Erdős Math Breakthrough 22:00 AI Displacement, UBI & the Eric Schmidt Backlash 27:00 Corporations as Governments & the Future of Local Economies 🔗 Links mentioned: OpenAI Erdős Solution: https://openai.com/index/model-disproves-discrete-geometry-conjecture/GitHub Hack: https://x.com/github/status/2056949168208552080 🐦 Follow the hosts: Pierson → https://x.com/piersonmarksBilal → https://x.com/deepwhitman 📺 More episodes: / @jellypodai

In Episode 44, Pierson and Bilal get into something every AI builder eventually runs into — the model is smart, but it still breaks. And it's not always the model's fault. They break down why agents fail not because they can't think, but because they have no eyes, arms, or legs — and what it actually takes to build systems where agents succeed. Pierson demos how he got agent browser to auto-authenticate behind a login wall using OAuth magic link flow, no password required. Bilal gets into Hyperframes, Remotion's parallelization edge, and the real cost of rendering long-form AI video at scale. From there they cover: Why you can outsource your thinking to AI, but never your understanding — an Andrej Karpathy quote worth sitting withThe Bun codebase port from Zig to Rust — a million lines of code, merged in daysToken maxing culture and why burning more tokens doesn't mean building betterJack Dorsey's Block org structure and why tooling engineers are quietly the most valuable people in the roomBilal releases an AI-generated song — and the comments didn't know it wasn't humanA guy who posted a real Monet and told the internet it was AI — and they believed itThe Chinese solo creator who made an AI zombie film in 10 days for $400 and got a studio deal Chapters: 00:00 Episode 44 & What We've Been Building This Week 06:00 Why AI Models Fail — And It's Not What You Think 12:00 Outsource Your Thinking, Not Your Understanding 19:00 Principal Engineers, Product Vision & What AI Can't Replace 27:00 Token Maxing, Tooling Teams & the Jack Dorsey Org Shift 33:00 Agent Auth, Playwright & Building Arms and Legs for AI 39:00 AI Art, the Monet Trick & Bilal's AI Song Nobody Knew Was AI 🔗 Links mentioned: Bilal's AI Song: https://x.com/deepwhitman/status/2055001909619274146Hyperframes: https://hyperframes.heygen.com/Bun Rewrite in Rust: https://news.ycombinator.com/item?id=48132488 🐦 Follow the hosts: Pierson → https://x.com/piersonmarksBilal → https://x.com/deepwhitman 📺 More episodes: / @jellypodai

In Episode 43, Pierson and Bilal come fresh off the a16z Generative Media Video Hackathon - and they built some wild stuff. Bilal went from a prompt to a full AI-generated podcast: custom avatars, Gemini TTS voices, lip-synced video, and live infographics pulled by deep research. All stitched together programmatically. Pierson built something different — an AI producer that joins your recording call as a guest, listens for a wake word, and pulls up images, search results, or visuals on a live canvas. Think: Jamie from Joe Rogan, but agentic. From there they get into: GPT Realtime 2 — real-time transcription, translation, and a voice API that stays listening even when "asleep"The shift from AI writing code to AI owning features — and what engineering responsibility looks like in that worldAI-generated anime breaking the last creative barrier humans thought they hadXAI dissolving into SpaceX, Elon's Colossus One deal with Anthropic, and what it means for Claude Code limitsMicro datacenters — Nvidia Blackwell chips mounted in your garage via SPAN's XFRA Chapters: 00:00 Episode 43 & The a16z Generative Media Hackathon 04:00 Bilal's Build: Prompt to Full AI Podcast with Avatars 10:30 Pierson's Build: An AI Producer That Joins Your Call Live 20:00 GPT Realtime 2 & the Always-On AI Future 31:00 AI Anime, Creativity & the Last Human Bastion 37:00 XAI, Anthropic's Colossus Deal & Micro Data centers 🔗 Links mentioned: GPT Realtime API: https://developers.openai.com/api/docs/models/gpt-realtimeSin Citium anime: https://x.com/Cont_animation/status/2051296715781619829SPAN Micro Datacenters x Nvidia: https://www.businesswire.com/news/home/20260414372626/en/SPAN-Announces-XFRA-a-Distributed-Data-Center-Solution-to-Close-the-Speed-to-Power-Gap-for-AI-Compute-Demand 🐦 Follow the hosts: Pierson → https://x.com/piersonmarksBilal → https://x.com/deepwhitman 📺 More episodes: / @jellypodai

In Episode 42 of Creative Flux, Pierson and Bilal kick off episode 42 - the answer to the universe - and things get weird fast. From there they cover: - ElevenLabs launching a standalone music social app, why Suno accidentally became a social network first, and the impossible position Spotify is now in - An AI podcast of Henrik Johansen and Martin Shkreli so realistic even Bilal got fooled - Talkie 13B — a model trained exclusively on pre-1931 text by Alec Radford, and why it feels like actually talking to someone from the past - Stripe Link going agent-first and Sam Altman hinting at something big around bring-your-own-tokens at Stripe Sessions Chapters: 00:00 Hitchhiker's Guide, Audio Books & the Future of Multi-Voice AI Narration 08:10 ElevenLabs Music, Suno's Social Network & What Spotify Is Really Up Against 16:00 Henrik Johansen x Martin Shkreli — The AI Podcast That Fooled Everyone 22:00 Talkie 13B — A Language Model Trained Exclusively on Pre-1930 Text 36:30 Stripe Link for Agents, Sign In with ChatGPT & the Agent-First Platform Shift 🔗 Show Notes: ElevenLabs Music: https://elevenlabs.io/music Henrick Johanssen & Martin Shkreli: https://x.com/compliantvc/status/2049535715369775559?s=12 Talkie 13B (pre-1930 LLM): https://talkie-lm.com/introducing-talkie Stripe Link for Agents: https://stripe.com/blog/giving-agents-the-ability-to-pay 🐦 Follow the hosts: Pierson → https://x.com/piersonmarks Bilal → https://x.com/deepwhitman 📺 More episodes: / @jellypodai

In Episode 41 of Creative Flux, Pierson and Bilal cover one of the most chaotic weeks in AI yet — GPT 5.5 literally dropped 20 minutes before they hit record. From there they cover: Opus 4.7's regression Anthropic finally admitted to, what actually broke in Claude Code, and the $25B Amazon deal behind their compute crunch The chip race — why Anthropic matching OpenAI on inferior hardware changes everything once they actually get Nvidia access Tidepool (formerly Claw Connect) — an open source peer-to-peer protocol letting isolated Claude Code instances talk to each other. Three agents, no shared context, they built a subscription business GPT Image 2's 300-point ELO jump and why it's an image agent, not just a model 🔗 Show Notes: Andon Store: https://x.com/jlagerros/status/2046966793538048295 Tidepool - Agent to Agent Communication Protocol: https://github.com/Jellypod-Inc/tidepool Opus 4.7: https://www.anthropic.com/news/claude-opus-4-7 GPT Image 2: https://developers.openai.com/api/docs/models/gpt-image-2 GPT 5.5: https://openai.com/index/introducing-gpt-5-5/ Hyperspace Pods: https://x.com/varun_mathur/status/2044882359565312468 Amit Jain from Luma AI on Unified Intelligence Systems (CS 153: Frontier Systems): https://www.youtube.com/watch?v=WNNrUuMQkl8 🐦 Follow the hosts: Pierson → https://x.com/piersonmarks Bilal → https://x.com/deepwhitman 📺 More episodes: / @jellypodai