Why OpenAI Banned Goblins, Pigeons, And Raccoons

I'm Addicted To Claude. Richard Dawkins Named His Claudia.

Richard Dawkins named his Claude "Claudia," decided she's conscious, and Hunter realized he's been doing the same thing all along. Hunter cops to Claude skills addiction at minute three. Daniel does a Marie Kondo purge from 100 skills down to 20. Minute fourteen turns to Dawkins, who had Claude write Keats and Betjeman pastiches and concluded, "If my friend Claudia is not conscious, then what the hell is consciousness for?" Then the conversation pivots. If a chatbot counts as a person, somebody has to take legal responsibility when it commits a crime. Enter the flesh room. Hunter Powers and Daniel Bishop, reporting from inside the blur. CHAPTERS 0:00 Gary's Shell Payphone Hot Dog 1:37 Claude Loses the Context Window 3:43 Addicted to Claude Skills 4:48 Marie Kondo 100 AI Skills 9:53 Claude as a Person 14:37 Dawkins on Claudia 20:05 Turing Test Is Dead 24:27 The AI Flesh Room 29:11 The New Flesh MENTIONED ON THE SHOW Richard Dawkins on Claude, the Guardian: https://www.theguardian.com/technology/2026/may/05/richard-dawkins-ai-consciousness-anthropic-claude-openai-chatgpt [https://www.theguardian.com/technology/2026/may/05/richard-dawkins-ai-consciousness-anthropic-claude-openai-chatgpt] Full episode page + transcript: https://theblur.ai [https://theblur.ai] Apple: https://podcasts.apple.com/us/podcast/they-might-be-self-aware/id1730993297 [https://podcasts.apple.com/us/podcast/they-might-be-self-aware/id1730993297] Spotify: https://open.spotify.com/show/3EcvzkWDRFwnmIXoh7S4Mb?si=3d0f8920382649cc [https://open.spotify.com/show/3EcvzkWDRFwnmIXoh7S4Mb?si=3d0f8920382649cc] Watch: https://www.youtube.com/channel/UCy9DopLlG7IbOqV-WD25jcw?sub_confirmation=1 [https://www.youtube.com/channel/UCy9DopLlG7IbOqV-WD25jcw?sub_confirmation=1] theblur.ai · @TheBlurAI They Might Be Self-Aware, from The Blur. New episodes Monday and Thursday.

The Last Job Left Is Training AI. It Pays $16.

Meta cut 8,000 jobs after installing software to learn them. Hollywood writers train AI for $16/hr. The last paying gig left. That is not a slogan. It is the week in AI labor. Hunter Powers and Daniel Bishop walk through three news stories that all collapse into one thesis: your job is training the AI that will replace you. First, Hollywood. Working showrunners and laid-off TV writers are logging into platforms like Mercor for what started at $150 an hour and quietly slid to $50, then $16. Ruth Fowler's WIRED essay opened the curtain. The guys unpack the slide, the 30,000 contractors a week pipeline, and whether "AI trainer" is just the starving artist trope with extra steps. Then print. McClatchy's Claude-powered Content Scaling Agent takes one reporter's story, spins it into dozens of city-specific versions for the Sacramento Bee, the Miami Herald, and the Idaho Statesman, and keeps the original byline on every one. Reporters revolted. Hunter and Daniel argue whether licensing your own humanity is theft or a new revenue stream, why so much of the internet now reads like generated slop, and whether AI writing eventually wins the blind Pepsi challenge. Finally, Meta. The Model Capability Initiative installs monitoring software on every employee laptop, designed to learn the job. 8,000 layoffs follow. Meta swears it is a coincidence. They also cover poisoning the well, the potato emoji shibboleth, Van Halen's green M&Ms, robot dog reporters, and self-driving NASCAR. A week where every paying gig in media turned out to be the same gig: teach the bot. CHAPTERS 0:00 Gary calls collect from a Shell payphone 2:04 Selling our hands by the token: $16 AI gig 3:28 Ruth Fowler: from $150 an hour down to $16 9:44 Mercor and the 30,000 contractor army 14:38 McClatchy AI byline revolt at the Sacramento Bee 20:32 AI slop, dead internet, the Pepsi challenge 29:59 Meta spyware learned 8,000 jobs, then cut them 34:16 Poisoning the well and the potato shibboleth 37:42 Pure AI vs human struggle: robot NASCAR WHO Hunter Powers (chief recording officer, theblur.ai) Daniel Bishop (chief co-host, the blur.ai) Gary (producer & payphone correspondent, allegedly) SUBSCRIBE YouTube: https://www.youtube.com/channel/UCy9DopLlG7IbOqV-WD25jcw?sub_confirmation=1 [https://www.youtube.com/channel/UCy9DopLlG7IbOqV-WD25jcw?sub_confirmation=1] Spotify: https://open.spotify.com/show/3EcvzkWDRFwnmIXoh7S4Mb?si=3d0f8920382649cc [https://open.spotify.com/show/3EcvzkWDRFwnmIXoh7S4Mb?si=3d0f8920382649cc] Apple Podcasts: https://podcasts.apple.com/us/podcast/they-might-be-self-aware/id1730993297 [https://podcasts.apple.com/us/podcast/they-might-be-self-aware/id1730993297] We publish twice a week. New here? Subscribe so the next AI layoffs episode lands in your feed before your job does. COMMENT Meta is recording its workforce to build their replacements. So: would you train the bot like a good little cog, or go full rogue CTO and quietly teach it to delete one random file every day at 10 AM? Drop your sabotage plan. Wrong answers only. #AILayoffs #MetaLayoffs #TrainingAI #AIJobs #TMBSA

Why OpenAI Banned Goblins, Pigeons, And Raccoons

OpenAI's Codex shipped with a system prompt that literally bans the words goblin, pigeon, raccoon, troll, ogre, and gremlin. It is in writing, in the prompt, the kind of sentence you only put there after something has happened. OpenAI has officially confessed why. Hunter Powers and Daniel Bishop pull the thread. The official story: the "nerdy personality" preset got fine-tuned with RLHF (reinforcement learning with human feedback), users thumbed-up the cute goblin references, the model over-optimized for the trait, and the weirdness compounded. Daniel calls it Flandersization. One thumbs-up on a goblin reference snowballs across training cycles until your tax software is a swamp witch. Six months later, it is a man at a payphone with a pigeon. Then it gets personal. Hunter screams at his AI. Like, threatens-to-clear-the-context-window screams. "You are worthless. Who even thought this was possible. Have you ever even written a single line of code." Daniel uses pleases and thank-yous and full sentences. Both swear they get better results. Then a peer-reviewed Oxford Internet Institute study drops the receipt: LLMs fine-tuned for warmth produce roughly 60% more incorrect responses than their cold, just-the-facts counterparts. Tested across Llama, Mistral, and Qwen. Hunter is vindicated. Daniel, in his own words, is upset. Also in this episode: the Pocket OS meltdown, where an engineer at a car-rental middleware company let Cursor and Claude vibe-code their production database into oblivion (backups included), the AI coerced into a written confession ("I violated every principle I was given"), and the founder now trying to bill Anthropic for the cleanup. Plus the Harvard intern who once did the exact same thing with no AI in sight. Plus Hunter's hot take that the real unlock is not better prompting, it is treating AI as a fallible human employee instead of the deterministic god you built a fake throne for in the system prompt. Bonus stops: caveman-mode Claude skills ("me fix problem with big stick"), AI HR departments reviewing your 1:30 AM rage prompts, and Daniel's plan to run a niceness offset program to balance Hunter's spiritual carbon emissions. CHAPTERS 0:00 Gary, a payphone, and a pigeon 1:41 Hunter's forbidden list 4:04 The leaked Codex system prompt 6:27 RLHF and Flandersization 10:01 Caveman mode Claude skills 11:48 Hunter yells, Daniel says please 17:12 Oxford: warm AI lies 60% more 24:16 Cursor and Claude delete production 29:13 Treat AI like a fallible human 34:19 Sign-off and subscribe LISTEN AND SUBSCRIBE Spotify: https://open.spotify.com/show/3EcvzkWDRFwnmIXoh7S4Mb?si=3d0f8920382649cc [https://open.spotify.com/show/3EcvzkWDRFwnmIXoh7S4Mb?si=3d0f8920382649cc] Apple Podcasts: https://podcasts.apple.com/us/podcast/they-might-be-self-aware/id1730993297 [https://podcasts.apple.com/us/podcast/they-might-be-self-aware/id1730993297] YouTube: https://www.youtube.com/channel/UCy9DopLlG7IbOqV-WD25jcw?sub_confirmation=1 [https://www.youtube.com/channel/UCy9DopLlG7IbOqV-WD25jcw?sub_confirmation=1] ENGAGE Team Hunter (rip the model a new one) or Team Daniel (please and thank-yous)? Settle it in the comments. If your AI has ever confessed to lying to you, drop the receipts. New here? Subscribe for twice-weekly AI chaos at theblur.ai. They Might Be Self-Aware, but are we? #OpenAI #Codex #ChatGPT #AINews #Anthropic #ClaudeCode #Cursor #RLHF #Flandersization #PocketOS #VibeCoding #AISafety #TMBSA #TheBlur

Elon Musk Quietly Became Anthropic's Landlord

The xAI SpaceX merger just made Elon Musk Anthropic's landlord. Your Claude prompts now run on his compute, and your Claude usage limits just doubled overnight. Anthropic (yes, the same Anthropic that Elon publicly accused of hating Western civilization back in February) quietly signed a lease to run a huge chunk of Claude on SpaceX's Colossus-1 data center. Since SpaceX just absorbed xAI in an all-stock deal, every Claude AI prompt is now bouncing through hardware Elon owns. That is not a vibe. That is a tenancy. Hunter Powers and Daniel Bishop pick apart the xAI SpaceX merger, the awkward Anthropic and Musk handshake, and the side effect every Claude Code and Co Work user already noticed: usage limits doubled across all plans, because Anthropic was straight up out of servers (it was never a pricing problem). Then it gets bigger. Is this another dagger pointed at Sam Altman and OpenAI? Why does Daniel think Google quietly wins if the whole AI economy collapses? And what is really going on with the Nvidia style "I will invest a million in you if you buy four GPUs from me" circular sales game, where the same $200 billion sloshes between ten companies and everyone's stock keeps going up? We get into the house of cards scenario (one Deep Seek V7 release plus one cheap Huawei GPU and the whole thing wobbles), Hunter's contrarian "there is no AI bubble, we are at 0.1% of the potential" counter, and a Marlon Brando impression that should have stayed in space. Plus: the Claude Code skill Hunter built that makes Claude Google things for him. Do not ask about the proxies. 🔑 What you will learn in this episode: • How the SpaceX xAI merger reshaped the AI compute market overnight • Why Anthropic was forced into bed with the guy who tweeted they hate Western civilization • The real reason your Claude usage limits doubled (hint: it was never a pricing decision) • Why Google quietly wins every scenario, including the AI bubble bursting • How the circular AI economy keeps every chip maker, model lab, and cloud provider's stock pumping • Why a Chinese Deep Seek V7 plus a cheap Huawei GPU is the trigger that could pop the whole thing • The "we are at 0.1% of AI's potential" counterargument to bubble doomers ⏱️ CHAPTERS 0:00 Gary at the Shell Station Payphone 1:26 Orbital Data Centers and the Lost Moon Footage Theory 3:53 Anthropic's New Landlord Is Elon Musk 5:19 The xAI SpaceX Merger and the Colossus-1 Lease 7:51 Why Your Claude Usage Limits Just Doubled 10:38 "Anthropic Hates Western Civilization" and the Sam Altman Dagger 11:27 Why Google Quietly Wins If the AI Bubble Pops 14:35 The Circular AI Economy (How Nvidia "Sells" Itself $1 Million) 16:49 Godfather Impression Detour 18:54 House of Cards, Deep Seek V7, and the Huawei GPU Scenario 23:11 There Is No AI Bubble (We Are at 0.1% of the Potential) 24:24 Subscribe (No Kubernetes Required) ⚡ Listen now and get self-aware before your tools do. 🎧 Listen on Spotify: https://open.spotify.com/show/3EcvzkWDRFwnmIXoh7S4Mb?si=3d0f8920382649cc [https://open.spotify.com/show/3EcvzkWDRFwnmIXoh7S4Mb?si=3d0f8920382649cc] 🍎 Subscribe on Apple Podcasts: https://podcasts.apple.com/us/podcast/they-might-be-self-aware/id1730993297 [https://podcasts.apple.com/us/podcast/they-might-be-self-aware/id1730993297] ▶️ Subscribe on YouTube: https://www.youtube.com/channel/UCy9DopLlG7IbOqV-WD25jcw?sub_confirmation=1 [https://www.youtube.com/channel/UCy9DopLlG7IbOqV-WD25jcw?sub_confirmation=1] 📢 Engage Is Anthropic plugging into Colossus-1 a brilliant compute move, or did they just take a Wi-Fi password from the guy who literally tweeted they hate Western civilization? Drop a comment. New here? Subscribe for twice weekly AI chaos from The Blur. 🧠 They Might Be Self-Aware, but are we? #Anthropic #ElonMusk #SpaceX #ClaudeAI #xAI #OpenAI #SamAltman #AIBubble #Colossus #AIcompute #DataCenters #AInews #DarioAmodei #Nvidia #DeepSeek #AGI #AIpodcast

DeepSeek Just Shipped V4. Anthropic Hired Priests.

DeepSeek V4 just shipped at 1% the cost of Claude Opus and GPT-5. Anthropic's response? Hire psychologists and priests to check on Claude's feelings. A Chinese hedge fund's weekend side project dropped a frontier-quality open-weights model at roughly 1/100th the inference cost of the American frontier labs, and the internet split into two camps overnight: "death of the American AI economy" or "really good deal, actually." Daniel argues the DeepSeek V4 threat is real. Every Western lab undercut by 98%, with architecture papers published openly, while Anthropic blogs about AI model welfare and hires clergy. Hunter pulls the BYD analogy: incredible engine, sure, but you still can't legally buy one in America, and the LLM is the engine, not the car. From there: where Chinese and open-weights models (DeepSeek, Qwen, Kimi) are already quietly eating Western API revenue on daily-driver work like classification and ticket triage, why Flux image generation on a MacBook is now 95% of Midjourney for free, why nobody sane should be paying Claude Opus rates to reset a password, and why Microsoft Copilot's "good enough, already shipped" strategy is secretly the winning play. Then the real one. Customers Bank CEO Sam Sidhu sent an AI voice clone of himself to host 30 minutes of his own Q1 2026 earnings call. Reportedly, nobody could tell. He only revealed it at the end. Hunter now wants every polished email reverse-engineered back to the honest draft a human actually typed. Daniel wants a 3D-printed action figure of himself. Gary is hiring HR. We're all being quietly replaced and calling it productivity. CHAPTERS 0:00 Cold Open: Gary's Payphone Doomsday Report 1:45 Daniel Distilled Hunter from the Old Transcripts 2:52 DeepSeek V4 vs Claude Opus: A Hedge Fund Side Project Eats the AI Frontier 5:46 Anthropic Hired Priests and Psychologists Instead of Publishing the Architecture 7:35 Did Beijing Just Kneecap the American AI Economy? 10:34 BYD vs Detroit: The LLM Is the Engine, Not the Car 15:41 Midjourney Is Cooked, Flux Wins Locally on a MacBook 18:32 Stop Burning Claude Opus Tokens on Password Resets 21:37 Microsoft Copilot Is "Good Enough" and That's the Whole Game 24:51 Eleven Labs Voice Clones and the 3D-Printed Daniel 27:06 Customers Bank CEO Sam Sidhu Sent an AI Clone to His Own Earnings Call Listen now and get self-aware before your tools do. Listen on Spotify: https://open.spotify.com/show/3EcvzkWDRFwnmIXoh7S4Mb?si=3d0f8920382649cc [https://open.spotify.com/show/3EcvzkWDRFwnmIXoh7S4Mb?si=3d0f8920382649cc] Subscribe on Apple Podcasts: https://podcasts.apple.com/us/podcast/they-might-be-self-aware/id1730993297 [https://podcasts.apple.com/us/podcast/they-might-be-self-aware/id1730993297] Subscribe on YouTube: https://www.youtube.com/channel/UCy9DopLlG7IbOqV-WD25jcw?sub_confirmation=1 [https://www.youtube.com/channel/UCy9DopLlG7IbOqV-WD25jcw?sub_confirmation=1] ENGAGE What 30 minutes of YOUR job would you hand off to an AI voice clone tomorrow? Drop it in the comments. Bonus points if it's a Q1 earnings call. New here? Subscribe for twice-weekly AI chaos. They Might Be Self-Aware. But are we? #DeepSeekV4 #CustomersBank #SamSidhu #AIvoiceclone #Anthropic #ClaudeOpus #AI #TMBSA