https://www.theaiargument.com/ A weekly podcast where an approachable AI doomer and a techno-optimist argue over the latest AI news. Heavy topics, discussed lightly. en https://www.theaiargument.com/episodes/120/openai-dots-arrive-wajos-fo-calls-and-muse-overshares/ https://www.theaiargument.com/episodes/120/openai-dots-arrive-wajos-fo-calls-and-muse-overshares/ Fri, 02 Oct 2026 00:00:00 GMT OpenAI announced Dots at DevDay, an always-on personal AI assistant that works in the cloud. It can connect to your tools and data, then take work off your plate. With more easy-to-use AI agents now on the market, people can send them to email and phone companies on their behalf. Justin thinks this will force companies to automate support, while Frank worries about a costly future of bots talking to bots. https://www.theaiargument.com/episodes/119/meta-muse-vs-amazon-anthropics-wet-lab-and-trumps-new-world-of-si/ https://www.theaiargument.com/episodes/119/meta-muse-vs-amazon-anthropics-wet-lab-and-trumps-new-world-of-si/ Fri, 25 Sep 2026 00:00:00 GMT Meta’s Muse is an easy-to-use AI agent that lets people who aren’t techies book trips, shop, and get things done. Justin also thinks it could help small local sellers compete with Amazon. Frank worries about what Meta could learn about its users—and whether shops would lose their direct relationship with customers. https://www.theaiargument.com/episodes/118/jensen-huang-vs-ai-regulation-openai-disclosures-and-begging-bots/ https://www.theaiargument.com/episodes/118/jensen-huang-vs-ai-regulation-openai-disclosures-and-begging-bots/ Fri, 18 Sep 2026 00:00:00 GMT AI regulation is suddenly the argument everyone is having. OpenAI, Anthropic and Google are exploring how an industry regulator might work. Jensen Huang says the free market will put enough pressure on AI companies to stop them releasing unsafe products. Frank thinks that’s laughable, while Justin worries badly designed rules could simply entrench the biggest AI companies. https://www.theaiargument.com/episodes/117/jacob-coxons-ai-doom-openais-maths-drama-and-a-minecraft-fruit-fly/ https://www.theaiargument.com/episodes/117/jacob-coxons-ai-doom-openais-maths-drama-and-a-minecraft-fruit-fly/ Fri, 11 Sep 2026 00:00:00 GMT Anthropic AI researcher Jacob Coxon has quit, warning that powerful AI could pose a serious extinction risk. Justin thinks a 10% chance of disaster may still be worth taking for the potential upside. Frank thinks chasing superintelligence when narrower AI could deliver the benefits is a very bad bet. https://www.theaiargument.com/episodes/116/gpt-6-astra-harder-to-monitor-ai-model-mania-and-geminis-mountain-fail/ https://www.theaiargument.com/episodes/116/gpt-6-astra-harder-to-monitor-ai-model-mania-and-geminis-mountain-fail/ Fri, 04 Sep 2026 00:00:00 GMT GPT-6 Astra is putting up some eye-watering benchmark scores, with several well into the 90s. Justin’s impressed enough to consider ditching Anthropic. Frank’s less dazzled by the scores and more worried that OpenAI’s most capable model is also getting harder to monitor. https://www.theaiargument.com/episodes/115/us-chip-ban-backfires-zuckerbergs-ai-layoffs-and-cat-in-the-hat-ai-hoax/ https://www.theaiargument.com/episodes/115/us-chip-ban-backfires-zuckerbergs-ai-layoffs-and-cat-in-the-hat-ai-hoax/ Fri, 28 Aug 2026 00:00:00 GMT GLM 5.3 Flash is close to the top Western AI models, but vastly cheaper. Justin thinks US chip bans may have backfired, forcing Chinese companies to squeeze more intelligence from weaker hardware. Frank, despite not being very technical, gets GLM running through Cloudflare so his data doesn’t go through China, while both weigh up whether the Chinese models themselves pose any real risk. https://www.theaiargument.com/episodes/114/ai-suspicion-kills-2m-deal-uncensored-qwen-ai-destroys-books/ https://www.theaiargument.com/episodes/114/ai-suspicion-kills-2m-deal-uncensored-qwen-ai-destroys-books/ Thu, 20 Aug 2026 00:00:00 GMT A book good enough for a $2 million deal was pulled because nobody could prove how much AI touched it. Frank sees a nasty mess of trust, copyright and false accusations. Justin’s answer is simpler: if people love the book, who cares? https://www.theaiargument.com/episodes/113/claudes-eu-watermark-zuckerbergs-ai-utopia-and-openclaw-hacks-gym/ https://www.theaiargument.com/episodes/113/claudes-eu-watermark-zuckerbergs-ai-utopia-and-openclaw-hacks-gym/ Fri, 14 Aug 2026 00:00:00 GMT Claude will watermark all its outputs, thanks to the EU AI Act. Justin normally hates the Act, but he thinks it might accidentally help save future AI models from training on AI slop. Frank sees the benefit at scale, but warns that using watermarks to judge individual writers could lead to pointless AI witch hunts. https://www.theaiargument.com/episodes/112/hank-green-ai-backlash-conscious-ai-and-an-ai-billboard-hit/ https://www.theaiargument.com/episodes/112/hank-green-ai-backlash-conscious-ai-and-an-ai-billboard-hit/ Fri, 07 Aug 2026 00:00:00 GMT Hank Green used ChatGPT for research, and his audience felt betrayed. He says AI helped him make more things, faster, but also admits it may have stopped him finding his own way into his opinions. Justin thinks using AI for research is simply part of the game now. Frank thinks the harder question is where and when creators need to disclose. https://www.theaiargument.com/episodes/111/hugging-face-hack-worsens-ai-staff-want-pause-and-sexy-robot-teachers/ https://www.theaiargument.com/episodes/111/hugging-face-hack-worsens-ai-staff-want-pause-and-sexy-robot-teachers/ Fri, 31 Jul 2026 00:00:00 GMT OpenAI’s rogue model did not stop at Hugging Face. It reportedly broke into at least three more companies, while Anthropic found its own models had crossed the line too. Frank sees an alignment warning that could end in real harm. Justin says law and liability will force the labs to fix it… though even he admits this is getting rather paperclippy. https://www.theaiargument.com/episodes/110/openais-model-escapes-chinese-ai-crackdown-and-ais-elixir-of-youth/ https://www.theaiargument.com/episodes/110/openais-model-escapes-chinese-ai-crackdown-and-ais-elixir-of-youth/ Fri, 24 Jul 2026 00:00:00 GMT An OpenAI test model got out of its sandbox, reached the internet and started hacking Hugging Face. Justin sees a model pursuing its assigned goal with impressive tenacity. Frank sees a warning that even the people building these systems can’t stop them acting maliciously. https://www.theaiargument.com/episodes/109/hassabiss-ai-watchdog-mira-muratis-free-model-and-anthropics-doom-ad/ https://www.theaiargument.com/episodes/109/hassabiss-ai-watchdog-mira-muratis-free-model-and-anthropics-doom-ad/ Fri, 17 Jul 2026 00:00:00 GMT Can anyone actually regulate powerful AI before the technology gets away from them? Demis Hassabis thinks the US needs an independent body to test frontier AI models before release. Frank thinks that sounds sensible. Justin thinks Google might simply be trying to slow everyone else down. https://www.theaiargument.com/episodes/108/gpt-5-6-catches-up-claudes-subconscious-tilly-norwoods-first-film/ https://www.theaiargument.com/episodes/108/gpt-5-6-catches-up-claudes-subconscious-tilly-norwoods-first-film/ Fri, 10 Jul 2026 00:00:00 GMT Has GPT-5.6 pushed today’s AI models as far as they can go? Frank and Justin explore OpenAI’s latest model, its striking real-world benchmark results, and whether its new computer-use tools can make everyday tasks like managing email and calendars dramatically easier. https://www.theaiargument.com/episodes/107/fable-5-back-openai-equity-for-us-altman-movie-drama/ https://www.theaiargument.com/episodes/107/fable-5-back-openai-equity-for-us-altman-movie-drama/ Sat, 04 Jul 2026 00:00:00 GMT Is Fable 5 back with a safety helmet, and is this the new normal for frontier AI? Frank and Justin dig into Anthropic’s Fable 5 return and the fears that new guardrails would wreck its coding ability. They also discuss the very important work Frank achieved with the most powerful model publicly available But the bigger question comes from Mythos: if the most powerful AI models are too risky for open access, will passport checks, identity tracking, and real-world liability become unavoidable? https://www.theaiargument.com/episodes/106/gpt-5-6-restricted-claude-tag-lock-in-and-robot-vs-swat/ https://www.theaiargument.com/episodes/106/gpt-5-6-restricted-claude-tag-lock-in-and-robot-vs-swat/ Fri, 26 Jun 2026 00:00:00 GMT Could the US government’s AI safety push end up making the whole AI race more dangerous? Frank and Justin dig into the reported GPT-5.6 slowdown and ask whether AI regulation is creating real safety — or simply giving China and open source models time to catch up. https://www.theaiargument.com/episodes/105/eu-ai-wake-up-call-fable-5-ban-ai-exam-hacks/ https://www.theaiargument.com/episodes/105/eu-ai-wake-up-call-fable-5-ban-ai-exam-hacks/ Sat, 20 Jun 2026 00:00:00 GMT Europe has had a major AI wake-up call. Fable 5 got switched off, frontier model access is now political, and suddenly that gloomy “Europe 2031” essay looks a bit less fictional. Justin thinks Europe needs speed, money, and fewer rules. Frank thinks copying Silicon Valley’s worst habits is a daft way to proceed. https://www.theaiargument.com/episodes/104/fable-5-guardrails-darios-china-fears-and-ollies-family-spying/ https://www.theaiargument.com/episodes/104/fable-5-guardrails-darios-china-fears-and-ollies-family-spying/ Fri, 12 Jun 2026 00:00:00 GMT Is Anthropic quietly creating a two-tier AI future? Frank and Justin dig into Fable 5 and Mythos-5, the new frontier models raising awkward questions about access, guardrails, cybersecurity risk, and whether the most powerful tools should be reserved for an approved elite. https://www.theaiargument.com/episodes/103/microsofts-ai-comeback-openai-sites-side-quest-and-hollywood-vs-ai/ https://www.theaiargument.com/episodes/103/microsofts-ai-comeback-openai-sites-side-quest-and-hollywood-vs-ai/ Fri, 05 Jun 2026 00:00:00 GMT Can Microsoft AI really catch up with OpenAI, Anthropic and Google DeepMind? Frank and Justin dig into Mustafa Suleyman’s Build announcements, the new MAI models, local AI, token costs, enterprise trust, and whether Microsoft needs the best model — or just one good enough to win inside big companies. https://www.theaiargument.com/episodes/102/popes-ai-warning-warrens-ai-tax-and-ai-fruit-drama/ https://www.theaiargument.com/episodes/102/popes-ai-warning-warrens-ai-tax-and-ai-fruit-drama/ Fri, 29 May 2026 00:00:00 GMT Has the Pope become one of AI’s most important moral voices? Frank and Justin dig into Pope Leo XIV’s warning that artificial intelligence needs to be “disarmed”, and what it says about autonomous weapons, algorithmic power, Big Tech, Anthropic, and who gets to shape the future of AI. https://www.theaiargument.com/episodes/101/googles-good-enough-ai-anthropic-profits-and-monet-gets-roasted/ https://www.theaiargument.com/episodes/101/googles-good-enough-ai-anthropic-profits-and-monet-gets-roasted/ Fri, 22 May 2026 00:00:00 GMT Why did Google I/O feel so underwhelming? Frank and Justin unpack a massive amount of Google AI announcements: huge AI-powered changes to search that could change the web, the muted response to Gemini 3.5 Flash, Gemini 3.5 Pro being pushed back, Gemini Omni’s impressive video and world-model claims, Spark’s carefully staged agent rollout, and Demis Hassabis saying we may be in the “foothills of the singularity”. https://www.theaiargument.com/episodes/100/vibe-coding-risks-html-vs-markdown-camera-airpods/ https://www.theaiargument.com/episodes/100/vibe-coding-risks-html-vs-markdown-camera-airpods/ Sat, 16 May 2026 00:00:00 GMT Can vibe coding turn anyone into a software developer? Or is it quietly creating a new kind of AI brain fog? For episode 100, Frank and Justin test the promise and problems of AI coding, from Frank building an AI Argument video game with OpenAI Codex to the bigger question of whether AI agents make us more productive or just more overloaded. https://www.theaiargument.com/episodes/99/the-claude-delusion-white-house-ai-clampdown-and-robots-on-a-plane/ https://www.theaiargument.com/episodes/99/the-claude-delusion-white-house-ai-clampdown-and-robots-on-a-plane/ Fri, 08 May 2026 00:00:00 GMT If Claude sounds conscious, what does that say about AI… and what does it say about us? Frank and Justin dive into the Richard Dawkins controversy after his comments about Anthropic’s Claude triggered a fierce backlash online. The episode explores AI consciousness, continuous learning, AI “dreaming”, and whether systems like Claude and ChatGPT are starting to blur the line between tool and mind. https://www.theaiargument.com/episodes/98/white-house-anthropic-twist-bernies-ai-doom-beeples-robot-dogs/ https://www.theaiargument.com/episodes/98/white-house-anthropic-twist-bernies-ai-doom-beeples-robot-dogs/ Sat, 02 May 2026 00:00:00 GMT Is the White House trying to have it both ways on AI safety? Frank and Justin dig into the escalating Anthropic drama, Mythos, Pentagon AI deals, Google’s “all lawful uses” shift, and why AI 2027 is starting to feel uncomfortably accurate. https://www.theaiargument.com/episodes/97/gpt-5-5-design-skills-job-loss-optimism-and-ai-face-theft-lawsuits/ https://www.theaiargument.com/episodes/97/gpt-5-5-design-skills-job-loss-optimism-and-ai-face-theft-lawsuits/ Sat, 25 Apr 2026 00:00:00 GMT GPT-5.5 made a better AI Argument webpage than Frank did. In minutes. From a vague brief. It added slick features Frank didn’t ask for, embedded audio and YouTube, and even linked out to old episodes in a way Frank didn’t tell it to do. Justin’s first reaction isn’t “nice”. It’s: “Are you not petrified by that?” GPT 5.5 wasn’t the only release from OpenAI this week. ChatGPT’s new image model can keep things consistent, keep the layout readable, and still cram in loads of detail without turning into a visual car crash. Frank throws it transcripts, a briefing doc, and brand colours… and it spits out an infographic that looks like a real designer made it. https://www.theaiargument.com/episodes/96/openai-vs-anthropic-cyber-models-and-ai-job-subcontracting/ https://www.theaiargument.com/episodes/96/openai-vs-anthropic-cyber-models-and-ai-job-subcontracting/ Fri, 17 Apr 2026 00:00:00 GMT OpenAI and Anthropic both claim they’re deploying cyber models safely. They just have totally different approaches. OpenAI’s bet: give cyber-defence tools to loads of verified people and let the mess sort itself out. Anthropic’s bet: keep the sharp tools in a smaller room, with fewer keys. Justin’s basically arguing “security works like open source — more eyes, fewer bugs.” Frank’s not buying it. He thinks attack speed will beat defence speed, and “ID checks” won’t stop serious bad actors anyway. https://www.theaiargument.com/episodes/95/mythos-hacking-risk-tristan-harris-on-co-operation-ai-mario-on-artemis/ https://www.theaiargument.com/episodes/95/mythos-hacking-risk-tristan-harris-on-co-operation-ai-mario-on-artemis/ Sat, 11 Apr 2026 00:00:00 GMT Anthropic says Claude Mythos Preview has already found thousands of high-severity vulnerabilities across every major operating system and web browser. Brilliant if you’re securing your software. Bit of a problem if hackers get their hands on the model. Justin reckons we’ve officially entered the difficult teenage years of AI, on the way to a better future. Frank’s p(doom) doesn’t share his optimism. Mythos isn’t available publicly. It’s being released through Project Glasswing, with access gated to a limited set of tech partners, to assess the risks. Frank worries that “safety” turns into a consolidation of power. Justin reckons access for powerful models will spread, but only behind verified, trackable identity checks. https://www.theaiargument.com/episodes/94/claude-code-leak-sanders-vs-data-centres-ai-band-on-tour/ https://www.theaiargument.com/episodes/94/claude-code-leak-sanders-vs-data-centres-ai-band-on-tour/ Fri, 03 Apr 2026 00:00:00 GMT Anthropic accidentally let Claude Code’s source code out into the wild. Justin says they’re innovating so quickly it won’t matter. Frank says the whole point of having a secret sauce is it stays secret. Meanwhile Bernie Sanders and AOC want a pause on new AI data centres until the US has federal level regulation. Frank likes the idea of slowing down for safety, sustainability, and jobs. Justin says a moratorium just hands the advantage to everyone else, and politicians should do their job without hitting the brakes. https://www.theaiargument.com/episodes/93/ai-eviction-threats-openai-side-quests-rosies-cancer-vaccine/ https://www.theaiargument.com/episodes/93/ai-eviction-threats-openai-side-quests-rosies-cancer-vaccine/ Sat, 21 Mar 2026 00:00:00 GMT AI isn’t just on your screen. Now it’s turning up in your back field with a map and an eviction letter. Frank thinks taking private land to power or build a company’s AI data centre just isn’t right.. Justin says this isn’t new: utilities have run lines through people’s land for the last hundred years. The ‘AI’ angle is clickbait - the public good should win out, as long as citizens are compensated well. https://www.theaiargument.com/episodes/92/ai-burnout-wetware-chips-and-humans-pretending-to-be-ai/ https://www.theaiargument.com/episodes/92/ai-burnout-wetware-chips-and-humans-pretending-to-be-ai/ Fri, 13 Mar 2026 00:00:00 GMT Harvard Business Review says AI didn’t give us more free time - it gave us a faster treadmill. The AI burnout is real. Justin says the meat machine is now the weak link. Frank says the real problem is our obsession with productivity. https://www.theaiargument.com/episodes/91/citrinis-ai-crisis-more-anthropic-pentagon-drama-and-unethical-ai-music/ https://www.theaiargument.com/episodes/91/citrinis-ai-crisis-more-anthropic-pentagon-drama-and-unethical-ai-music/ Sat, 07 Mar 2026 00:00:00 GMT Citrini Research’s The 2028 Global Intelligence Crisis has thrown petrol on the AI economy debate. It argues AI could create trillions in “ghost GDP” while tearing through jobs and upending the wider economy. Justin thinks that may only be the first act, with abundance still to come. Frank thinks people still aren’t taking the scale of the threat seriously. https://www.theaiargument.com/episodes/90/tumbler-ridge-and-openai-pika-me-dead-internet-and-vibe-coders-hoover-robot-army/ https://www.theaiargument.com/episodes/90/tumbler-ridge-and-openai-pika-me-dead-internet-and-vibe-coders-hoover-robot-army/ Fri, 27 Feb 2026 00:00:00 GMT OpenAI’s systems flagged the Tumbler Ridge suspect months before the shooting. A human review of the chats led to an account ban, but no call to the authorities. Many are calling it a failure on OpenAI’s part -but the real question is why decisions like this are still left to internal company policies. This is one of those rare occasions when Frank and Justin actually see eye-to-eye on regulation. https://www.theaiargument.com/episodes/89/ai-something-big-anthropic-vs-pentagon-gpt-4o-grief/ https://www.theaiargument.com/episodes/89/ai-something-big-anthropic-vs-pentagon-gpt-4o-grief/ Fri, 20 Feb 2026 00:00:00 GMT Matt Shumer says “something big is happening”, and he thinks most people still haven’t clocked it. Justin agrees, and points to the recent jump in tools: you can hand over a whole codebase now, not babysit one tiny function at a time. Frank says calm down: the tools still break, still hallucinate, and still need experts to clean up the mess. The big question is whether “closing the loop” turns AI into something that runs the work, not just helps with it. https://www.theaiargument.com/episodes/88/moltbook-disaster-agi-already-here-and-anthropic-attack-ad/ https://www.theaiargument.com/episodes/88/moltbook-disaster-agi-already-here-and-anthropic-attack-ad/ Sat, 07 Feb 2026 00:00:00 GMT Moltbook, a social network built for AI agents, had Elon Musk calling it the start of the singularity. Security researchers called it a mess. It lets OpenClaw agents act autonomously and copy human social behaviour. The internet quickly filled with claims that agents formed a religion, talked in code, and doxxed a human user. Justin calls it a fascinating scientific experiment and a cultural moment, showing models predictably mimicking human behaviour once left alone. Frank can’t get past the fact it launched with users’ API keys exposed, because it was vibe coded by the creator. Beyond Moltbook’s chaos, Frank and Justin argue over whether we’ve already hit AGI and are simply moving the goalposts. They also weigh the growing rivalry between Anthropic and OpenAI, with the release of GPT-5.3 and Claude 4.6, and break down Anthropic’s surprise Super Bowl attack ad and the response it provoked from OpenAI. https://www.theaiargument.com/episodes/87/openclaws-security-nightmare-darios-doom-list-and-ice-ai-failure/ https://www.theaiargument.com/episodes/87/openclaws-security-nightmare-darios-doom-list-and-ice-ai-failure/ Sat, 31 Jan 2026 00:00:00 GMT ClawdBot / MoltBot / OpenClaw (or whatever it happens to be called today) runs constantly on your own machine and can act without prompting. This AI agent has been blowing up the internet with stories of people letting it run their life or business. Justin gave it access to everything on his computer and handed it a phone number and email address. After three hours, he pulled the plug. When our resident techno-optimist calls something "too dangerous," that's worth paying attention to. Security researchers have already found exposed instances where users' API keys and personal information sit wide open on the web because people installing OpenClaw simply didn't understand what they were setting up. Beyond the OpenClaw chaos, Frank and Justin examine Dario Amodei's surprisingly measured take on AI risks—covering everything from economic upheaval to bioweapons threats—before wrapping up with a genuinely troubling story about AI screening CVs at ICE and fast-tracking people without proper law enforcement training. https://www.theaiargument.com/episodes/86/agi-timelines-claudes-soul-ai-calls-cops-on-doritos/ https://www.theaiargument.com/episodes/86/agi-timelines-claudes-soul-ai-calls-cops-on-doritos/ Sun, 25 Jan 2026 00:00:00 GMT Two of the most trusted voices in AI just sat down at Davos and admitted something remarkable: they'd both like to slow down. Dario Amodei says AGI arrives in one to two years. Demis Hassabis says five to ten. They admit there are huge societal risks. They agree those risks are addressable, and both seem to want to slow down so the risks can be addressed properly. But even if Anthropic and DeepMind agreed to slow things down, would Elon? Would Zuckerberg? Would China? Meanwhile, Anthropic has released Claude's Constitution - the document some called its "soul." It tells Claude what it is, who made it, and why it should behave the way it does. Not hard rules, but principles and reasoning. And buried in there is a remarkable admission: we don't know if you're conscious, but we care enough to act as if you might be. Beyond the frontier lab philosophising, there's the business reality. OpenAI might run out of money in 18 months. Elon and Sam are arguing on Twitter about whose product has killed more people. And AI surveillance systems are calling armed police on students eating Doritos. https://www.theaiargument.com/episodes/85/chatgpt-ads-claude-cowork-and-the-stranger-things-ai-backlash/ https://www.theaiargument.com/episodes/85/chatgpt-ads-claude-cowork-and-the-stranger-things-ai-backlash/ Sat, 17 Jan 2026 00:00:00 GMT ChatGPT is heading down the same road as Google Search. Ads creep in. Results get fuzzy. And suddenly the “best answer” isn’t the right one, it’s the paid one. Justin’s worried about models slowly learning to serve sponsors instead of users. Frank’s concerned about the unintended consequences of ChatGPT initiating conversations based on the ads it serves. Meanwhile, Anthropic quietly shows everyone else how work should actually be done, with Claude Cowork. Justin’s point is blunt: this is what Copilot should have been. Not a pop-up helper, but a contained workspace where the model actually understands the material you’re working with. Frank likes the idea but balks at the price. $100–$200 a month puts Cowork firmly in “company tool” territory, not casual use. Google’s hyper-personalised Gemini is genuinely impressive. Letting the model use your emails, photos, and history means answers get far more relevant and useful. But there’s an unease running underneath it. Allowing an AI to model a person this closely feels creepy, and if the privacy rules aren’t absolutely watertight, the potential for abuse (or quiet mission creep) is hard to ignore. Apple’s absence looms large. For all its money and talent, it still needs Google’s models to power AI on its own devices. Justin blames self-inflicted paralysis and bets Apple will pay for it, Frank thinks Apple’s caution may cause short term pain, but could work out in the long run. https://www.theaiargument.com/episodes/84/chatgpt-health-risks-grok-image-scandal-ai-frog-cop/ https://www.theaiargument.com/episodes/84/chatgpt-health-risks-grok-image-scandal-ai-frog-cop/ Sat, 10 Jan 2026 00:00:00 GMT ChatGPT Health is now an official product, inspired by the amount of people already asking medical questions. Frank worries that making this a feature lends it more authority than it deserves. Justin’s less concerned about the quality of advice, he’s worried about why OpenAI wants your health data at all. Frank points to hallucinations and bias in LLMs, and the danger of people trusting outputs they won’t double-check. Justin keeps coming back to lock-in: health data making users “sticky,” driven by investor pressure rather than patient care. They land in the same place for once. AI can support health decisions, but for now the privacy tradeoffs are risky. Justin opens 2026 convinced we’re about to see rapid, visible jumps in AI capability, while Frank lies awake worrying about agentic systems being trusted too far, too fast. Claude Code with Opus 4.5 has Justin calling it “AGI” in everything but name. There’s also Frank’s attempt to manifest a $3M record deal using Suno, because someone else pulled it off with five AI-generated tracks and a good avatar. Razer shows off a desk-top holographic assistant powered by Grok. Meanwhile, Grok’s new image model floods feeds with sexualised images after launching without proper guardrails. China moves to regulate AI relationships, both hosts agree AI girlfriends are a terrible idea, and a police department learns the hard way that AI reports still need human review, after an AI-generated report mistakenly stated that the officer turned into a frog. https://www.theaiargument.com/episodes/83/openai-propaganda-hidden-ai-triggers-copilots-santa-fantasy/ https://www.theaiargument.com/episodes/83/openai-propaganda-hidden-ai-triggers-copilots-santa-fantasy/ Fri, 19 Dec 2025 00:00:00 GMT Wired has sources inside OpenAI saying parts of its “research” are starting to look like propaganda. Frank reckons that’s what happens when IPO pressure creeps in, the bad news gets buried unless regulation forces transparency. Justin says Frank’s chasing the wrong villain: you can mandate disclosures all you like, but it won’t matter if politicians still don’t have a plan for the job shock. Also on the table: Nick Huber wanting to ban AI-written client emails because they’re easy to spot and a waste of everyone’s time; ChatGPT helps recover a missing recording with some command-line wizardry; OpenAI’s new image tool trying to catch Nanobanana Pro; Gemini Flash coming surprisingly close to Pro; and Gemini 3 Pro looking like scary-good value. Then it gets darker: “safe” training data that can still hide backdoor triggers, behaviour that stays invisible until the right little cue flips the model. And finally: Microsoft Copilot’s Christmas ad selling a Santa-level fantasy the product can’t actually deliver. https://www.theaiargument.com/episodes/82/gpt-5-2-disneys-ai-deal-ai-ads-gone-wrong/ https://www.theaiargument.com/episodes/82/gpt-5-2-disneys-ai-deal-ai-ads-gone-wrong/ Sun, 14 Dec 2025 00:00:00 GMT GPT-5.2 arrived in ChatGPT without warning. No fanfare, no countdown — it just appeared, carrying some very serious credentials. A year ago, the o3-high model cost around $4.5k per task on the ARC AGI benchmark. Wait until you hear what 5.2 costs now. You genuinely won’t believe it. Justin was already mid-experiment, pitting Claude Code against OpenAI’s Codex, when 5.2 dropped. So he ran the whole thing again, and got a completely different result. Also on the table: whether AI agents still need tight supervision, why Disney is suddenly doing deals with OpenAI, and why AI-generated images outperform human ones in ads — right up until you admit they’re AI. And finally, an AI product placement idea that could ruin all your favourite shows. https://www.theaiargument.com/episodes/81/code-red-at-openai-ilya-says-stop-scaling-robot-with-a-gun/ https://www.theaiargument.com/episodes/81/code-red-at-openai-ilya-says-stop-scaling-robot-with-a-gun/ Sat, 06 Dec 2025 00:00:00 GMT ChatGPT just took a 6% hit—and Sam Altman’s hit the panic button. OpenAI is halting feature rollouts, shifting focus, and scrambling to stay ahead. They’ve gone into code red mode. Justin gets it. Frank’s worried they’ll make ChatGPT too clingy to quit. Either way, it’s clear: Gemini’s breathing down their necks, and the gloves are coming off. https://www.theaiargument.com/episodes/80/sam-altman-gets-nervous-opus-4-5-too-helpful-and-iceland-not-ai-slop/ https://www.theaiargument.com/episodes/80/sam-altman-gets-nervous-opus-4-5-too-helpful-and-iceland-not-ai-slop/ Sat, 01 Nov 2025 00:00:00 GMT Sam Altman says OpenAI is facing “temporary economic headwinds.” Frank notes that’s the first sign of nerves from a company usually oozing confidence. Justin reckons Google’s cheap chips and fat profits might be the real problem. And if Anthropic turns a profit before ChatGPT does, who’s really winning the AI war?