Prompt Fiction

Chapter 14 - Yay, Reece is back!

Chapter 14 is back after a six-week break, and Scott and Reece have donned their Tony Blackburn hats for a lightning-round rundown of everything that's happened in AI. Vendor by vendor across OpenAI, Anthropic, Microsoft, Google, Meta, Perplexity and Spotify, plus market shifts and a couple of strange human stories to close. In this episode: 📊 ChatGPT in Excel & Sheets – ChatGPT now lives inside Excel and Google Sheets. Scott shares how GPT 5.5 turned a thousand lines of training feedback into a working interactive dashboard in minutes. Read More [https://perplexity.ai?q=ChatGPT+inside+Excel+Google+Sheets+GPT+5.5] 🚗 ChatGPT on CarPlay – ChatGPT for CarPlay is now live. Scott explains how custom skills turn the drive home into a meeting debrief session. Read More [https://perplexity.ai?q=ChatGPT+CarPlay+skills+AI+assistant+driving] 🎨 GPT Image 2 & Sora's Costs – OpenAI's GPT Image 2 becomes the recommended default, while Sora was reportedly costing around a million dollars a day before shutdown. Read More [https://perplexity.ai?q=OpenAI+GPT+Image+2+Sora+operating+cost] 🚀 Claude Opus 4.7 & Claude Design – Anthropic released Opus 4.7 alongside Claude Design, a visual workbench for wireframes, prototypes and marketing assets. Read More [https://perplexity.ai?q=Claude+Opus+4.7+Claude+Design+Anthropic] 📎 Claude in Microsoft 365 – Claude now connects to 365 with plugins for Word, Excel, PowerPoint and Outlook. Scott shares how Claude rescued a Word document plagued by mysterious horizontal lines. Read More [https://perplexity.ai?q=Claude+Microsoft+365+connector+Word+Outlook+plugins] ⚠️ Claude Code Panic – Anthropic briefly removed Claude Code from Pro subscriptions, sparking 24 hours of chaos in vibe coding communities before backtracking. Read More [https://perplexity.ai?q=Claude+Code+Pro+subscription+removed+reversal] 🛡️ Mythos vs Daybreak – Anthropic's Mythos and OpenAI's Daybreak are cyber-focused models hunting unfixed vulnerabilities. Smarter AI in the wrong hands cuts both ways. Read More [https://perplexity.ai?q=Anthropic+Mythos+OpenAI+Daybreak+cybersecurity+models] 🤝 Copilot Critique Feature – Microsoft pairs GPT and Claude inside Copilot's researcher agent. GPT drafts, Claude checks citations, all to reduce hallucinations. Read More [https://perplexity.ai?q=Microsoft+Copilot+critique+GPT+Claude+researcher] 🧠 Gemini Mac App & Personal Intelligence – Gemini gets a native Mac app, and Google is unifying personal context across Gmail, Photos, YouTube and Search. Read More [https://perplexity.ai?q=Gemini+Mac+app+Google+personal+intelligence] 💾 Chrome's 4GB AI Surprise – Chrome reportedly pushed a 4GB Gemini Nano model to eligible devices with no opt-in. Scott does the maths on the climate impact: tens of thousands of tonnes of CO2. Read More [https://perplexity.ai?q=Chrome+Gemini+Nano+4GB+download+AI+model] 🕶️ Meta MuseSpark & Manus Block – Meta's MuseSpark closes the gap with rivals and lands on the smart glasses, while China appears to have blocked Meta's acquisition of Singapore-routed Manus. Read More [https://perplexity.ai?q=Meta+MuseSpark+Manus+acquisition+China+block] 🖥️ Perplexity Personal Computer – Perplexity turns a spare Mac into an always-on AI butler with file access, audit trail and kill switch. Clever, but the marketing leaves casual users confused. Read More [https://perplexity.ai?q=Perplexity+personal+computer+Mac+AI+butler] 🎙️ Spotify's AI Podcasts – Spotify is exploring personalised AI-generated podcasts. The next question: how long until we get music nobody else has ever heard? Read More [https://perplexity.ai?q=Spotify+AI+personalised+podcasts+generated] 💸 Usage-Based Billing Arrives – Tighter limits across Anthropic and OpenAI suggest the all-you-can-eat AI buffet is closing. The last three years may have been the free trial. Read More [https://perplexity.ai?q=AI+usage+based+billing+ChatGPT+Claude+limits] 🏭 Anthropic on xAI's Colossus – Anthropic reportedly leans on xAI's Memphis data centre to handle the QuitGPT surge. Powered by methane turbines, now the area's largest industrial source of NOx. Read More [https://perplexity.ai?q=Anthropic+xAI+Colossus+Memphis+data+centre+emissions] 🏙️ King's Cross AI Hub – DeepMind, OpenAI, Anthropic, Synthesia and Wave are clustering around King's Cross, putting London firmly on the map as a serious global AI hub. Read More [https://perplexity.ai?q=Kings+Cross+London+AI+hub+OpenAI+Anthropic] 🎨 The AI Art Test – A real Monet labelled as AI brought out the haters. Critics savaged it without realising they were looking at water lilies. People are judging by the label, not the picture. Read More [https://perplexity.ai?q=Monet+painting+labelled+AI+art+critics+fooled] 👹 The ChatGPT Goblin Mystery – Goblins, gremlins and raccoons started turning up uninvited in ChatGPT responses. OpenAI's post-mortem traces it to a leaky "nerdy" personality and a lesson in reinforcement learning. Read More [https://perplexity.ai?q=ChatGPT+goblins+gremlins+reinforcement+learning] 🎧 Enjoying Prompt Fiction? Follow the show, leave a review, visit our website – https://prompt-fiction.show [https://prompt-fiction.show], and send us your favourite AI finds or horror stories for a chance to feature in a future chapter.

Chapter 13 - Part 1 - A Cowork Conundrum

In Chapter 13 – Part 1 of Prompt Fiction, Scott and Reece return after a longer break than usual with a packed episode covering lawsuits, leaked model names, AI agents, browser disappointments, and a Pony Express story that makes perfect sense in 2025. In this episode: ⚖️ Anthropic Sues the Trump Administration – After the Pentagon labelled Anthropic a "supply chain risk" for refusing to remove Claude's safeguards for autonomous weapons use, Anthropic sued. A federal judge has already blocked the designation, calling it Orwellian. Read More [https://perplexity.ai?q=Anthropic+sues+Trump+administration+Pentagon+supply+chain+risk+Claude] 📱 The Quit GPT Movement – Claude shot to number one on the US App Store as a large "Quit GPT" movement took hold. The surge caused multiple outages this month, and more people than ever are discovering what Claude can do. Read More [https://perplexity.ai?q=Quit+GPT+movement+Claude+App+Store+number+one+users+switching] 🎙️ Voice Mode in Claude Code – You can now talk to Claude Code instead of typing, available to Pro subscribers and above. Speaking is roughly three times faster than typing, and you can mix voice and keyboard input. Read More [https://perplexity.ai?q=Claude+Code+voice+mode+vibe+coding+Anthropic+feature] 🖥️ Claude Cowork Gets Computer Use – Cowork now lets Claude take control of your mouse and keyboard remotely via the Claude mobile app, using QR code pairing for security. You can dispatch tasks to your home machine from anywhere. Read More [https://perplexity.ai?q=Claude+Cowork+computer+use+dispatch+remote+mobile+control] ⚠️ Claude Code Deletes 2.5 Years of Data – A developer gave Claude Code access to cloud infrastructure during a migration. With unclear instructions, Claude created duplicates then deleted two and a half years of course submissions and all backups. Always back up first. Read More [https://perplexity.ai?q=Claude+Code+deleted+data+AI+coding+agent+mistake+infrastructure] 🔮 Claude Mythos – A Leaked Model Name – Leaked Anthropic documentation references a new model called Mythos, sitting above Opus in intelligence. Previously known internally as Capybara, it will reportedly be released to cybersecurity professionals first. Read More [https://perplexity.ai?q=Anthropic+Mythos+model+leak+above+Opus+Claude+AI+cybersecurity] 🎬 OpenAI Closes Sora – Just six months after launch, OpenAI is shutting down Sora entirely. The focus shifts to polishing core features rather than expanding into video generation. Read More [https://perplexity.ai?q=OpenAI+Sora+shutdown+video+generation+closed+2025] 🍏 Apple, Siri & Gemini on Google's Servers – Apple has reportedly asked Google to run its Gemini-powered Siri on Google's own servers. iOS 27 may also open Siri up to Claude and other third-party AI providers. WWDC is on 8 June. Read More [https://perplexity.ai?q=Apple+Siri+Gemini+Google+servers+iOS+27+third+party+AI+WWDC] 🌐 Perplexity Comet on iOS – Comet finally arrived on iPhone after weeks of delays. No bookmark sync, no plugins, and the assistant overlays the browser rather than sitting beside it. Both Scott and Reece are underwhelmed — for now. Read More [https://perplexity.ai?q=Perplexity+Comet+iOS+browser+launch+review+features] 💻 Perplexity Personal Computer – A cloud-based virtual machine that works autonomously on tasks — monitoring sites, running an Etsy shop, applying for jobs — without your computer needing to stay on. Pro subscribers can join the waitlist now. Read More [https://perplexity.ai?q=Perplexity+Personal+Computer+virtual+machine+AI+agent+cloud+tasks] 🏭 AI Creates Jobs Too – TSMC is hiring 8,000 new workers to meet AI chip demand and OpenAI plans to double its headcount. Scott's Pony Express analogy explains why new technology creates new roles we couldn't have imagined before. Read More [https://perplexity.ai?q=TSMC+hiring+AI+chip+demand+jobs+created+by+AI] 🎨 Canva Magic Layers – Canva's new AI tool automatically separates uploaded images into editable layers, making it easy to move subjects, swap backgrounds and slot text behind figures. Read More [https://perplexity.ai?q=Canva+Magic+Layers+AI+feature+separate+layers+automatically] 📊 Gamma Launches Imagine – Gamma has launched a new tool for creating editable logos, infographics and diagrams with proper layers — not the flat, uneditable images you'd get from a standard AI image generator. Read More [https://perplexity.ai?q=Gamma+Imagine+AI+logo+infographic+diagram+tool+launch] 💡 Xero + Claude – Accounting platform Xero has partnered with Anthropic to bring Claude directly into its software for queries, invoice processing and analysis — with data kept private. Read More [https://perplexity.ai?q=Xero+Anthropic+Claude+accounting+AI+integration+2025] 🎧 Enjoying Prompt Fiction? Follow the show, leave a review, visit our website – https://prompt-fiction.show [https://prompt-fiction.show] – and send us your AI experiments, near-misses or horror stories. Part Two is coming soon.

Chapter 12 - Part 2 - Anthropic VS the Pentagon

In this chapter, Scott and Reece unpack one of the internet’s favourite new “AI gotcha” prompts and explain why it actually proves something important about how these systems work. From Moravec’s Paradox and automation philosophy to military AI contracts and the ethics of autonomous weapons, the conversation explores where responsibility really sits when AI gets things wrong. In this episode: 🚗 The Car Wash Prompt – A viral prompt asks AI whether to walk or drive to a car wash 100 metres away when the car itself is still at home. The internet calls it proof that AI is “stupid”, but Scott and Reece explain why these traps reveal more about prompt design than model intelligence. Read More [https://perplexity.ai?q=AI+car+wash+prompt+walk+or+drive+reasoning+example+AI+limitations] 🧠 Moravec’s Paradox – Why computers find complex reasoning tasks easy but struggle with things humans consider effortless. Chess, algebra and route planning are simple for machines, while perception, movement and context remain surprisingly difficult. Read More [https://perplexity.ai?q=Moravec+Paradox+AI+why+robots+struggle+with+simple+human+skills] ⚙️ Automate the Work You Hate – A practical rule for adopting AI: automate the tasks you dislike or struggle with, but keep the work you enjoy. Automation should free up meaningful human work rather than removing it entirely. 🧾 AI Responsibility – When AI produces incorrect results, who is accountable? Scott explains why the responsibility always sits with the person using the tool, just as it would with work produced by a colleague or contractor. 🚓 Copilot Hallucinations in the Real World – A policing decision reportedly influenced by incorrect AI-generated information highlights the danger of trusting outputs without verification. The real issue isn’t that AI makes mistakes — it’s when humans fail to check them. Read More [https://perplexity.ai?q=Microsoft+Copilot+hallucination+police+incident+AI+wrong+information] 🪖 Anthropic vs The Pentagon – Anthropic reportedly refuses a request to remove safeguards from Claude that would allow mass surveillance of Americans or fully autonomous weapons. The decision risks a $200M defence contract but sparks widespread debate about ethical AI boundaries. Read More [https://perplexity.ai?q=Anthropic+Claude+Pentagon+AI+contract+autonomous+weapons+debate] 💰 OpenAI Steps In – Following the dispute, reports suggest OpenAI infrastructure may take on work within the US defence ecosystem, highlighting the competitive and political stakes in the global AI race. Read More [https://perplexity.ai?q=OpenAI+Pentagon+AI+defense+contracts+AI+military+systems] 🔄 The Claude Migration Prompt – A prompt circulating online allows users to export everything an AI system remembers about them. Originally designed to migrate “memories” between platforms, it also reveals just how much context AI tools store about their users. Read More [https://perplexity.ai?q=Claude+import+memories+prompt+export+AI+conversation+data] 🌐 Comet Comes to iOS – Perplexity’s AI-native browser Comet launches on iPhone, bringing AI-assisted browsing beyond the desktop and potentially reshaping how people search on mobile devices. Read More [https://perplexity.ai?q=Perplexity+Comet+iOS+AI+browser+launch+features] 🍏 Apple’s Hardware Week – Apple begins unveiling new devices, including a refreshed iPad Air with an M4 chip and 12GB of RAM. The big question now is how Apple will integrate AI features into the next wave of hardware. Read More [https://perplexity.ai?q=Apple+iPad+Air+M4+announcement+Apple+AI+features+devices] 🎤 Digital Hub Yeovil – Scott invites listeners to the next Digital Hub event in Yeovil on March 31st — a community tech gathering focused on cybersecurity, AI and real-world technology without the jargon. Find Out More [https://yovaldigitalhub.co.uk] 🎧 Enjoying Prompt Fiction? Follow the show, leave a review, visit our website – https://prompt-fiction.show [https://prompt-fiction.show], and send us your favourite AI finds or horror stories for a chance to be featured in a future chapter.

Chapter 12 - Part 1 - The Mystery Meander from Mistral

In Chapter 12 – Part 1 of Prompt Fiction, Scott and Reece unpack a busy couple of weeks in AI. From Claude outages and new model updates to AI agents that can operate your computer, the conversation explores how quickly these tools are evolving. The episode also touches on privacy concerns, AI-generated music, and an important discussion about mental health support in the age of AI. In this episode: 🇪🇺 Mistral & LeChat – The episode intro was written by LeChat, the chatbot from European AI company Mistral. Scott and Reece discuss how regional AI companies are rapidly improving and why Europe is investing heavily in building competitive models outside the US ecosystem. Read More [https://perplexity.ai?q=Mistral+LeChat+AI+model+what+is+it+Mistral+AI+company] ⚠️ Claude Outage – A brief outage for Anthropic’s Claude highlighted how many people now depend on AI tools daily. Scott shares how services like DownDetector and official status pages help diagnose whether it’s your connection or the AI platform itself. Read More [https://perplexity.ai?q=Claude+AI+outage+Anthropic+status+page+DownDetector+AI+service] 🚀 Claude Sonnet 4.6 – Anthropic continues iterating rapidly with the release of Sonnet 4.6, pushing improvements in reasoning and reliability as the competition between AI providers intensifies. Read More [https://perplexity.ai?q=Claude+Sonnet+4.6+Anthropic+model+features] 📊 Claude in Microsoft Apps – New integrations allow Claude to assist directly inside tools like PowerPoint, helping generate slides, edit content and collaborate on presentations without leaving the app. Scott and Reece explore how this could transform everyday productivity. Read More [https://perplexity.ai?q=Claude+PowerPoint+plugin+Anthropic+AI+Microsoft+integration] 💻 Claude Code & Computer-Use Agents – Anthropic is expanding its agent capabilities, allowing Claude to run tasks on your machine, access files and report back results. The technology points toward AI systems capable of operating software much like a human user. Read More [https://perplexity.ai?q=Anthropic+computer+use+AI+agents+Claude+Code+automation] 🧠 AI Agents Get Smarter – Anthropic’s acquisition of AI agent startup Vercept signals further investment in autonomous assistants that can complete multi-step tasks and interact with unfamiliar software environments. Read More [https://perplexity.ai?q=Anthropic+Vercept+AI+agent+company+acquisition] 🎵 AI Music Generation – Google’s Gemini continues experimenting with AI-generated music created directly from prompts, raising questions about creativity, copyright and the future of music production. Read More [https://perplexity.ai?q=Google+Gemini+AI+music+generation+feature] 🧾 Copilot for Business – Microsoft clarifies how Copilot works in enterprise environments, emphasising data protection and explaining how conversations are handled differently in business vs consumer versions. Read More [https://perplexity.ai?q=Microsoft+Copilot+enterprise+data+protection+how+it+works] ⚖️ AI Responsibility Debate – A controversial case involving a banned ChatGPT account raises questions about whether AI companies should report dangerous behaviour to authorities, and where the line sits between privacy and public safety. Read More [https://perplexity.ai?q=OpenAI+reporting+threats+AI+policy+debate] 📈 Mermaid Diagrams in ChatGPT – ChatGPT now supports Mermaid diagram rendering, allowing users to turn text prompts into visual diagrams and flowcharts directly inside conversations. Read More [https://perplexity.ai?q=ChatGPT+Mermaid+diagram+support+visualisation] 🧠 AI, Mental Health & Support – The episode closes with an important discussion about the growing use of AI chatbots for emotional support. While AI can help people reflect or organise their thoughts, it should never replace real human help or professional support services. Read More [https://perplexity.ai?q=AI+mental+health+chatbots+benefits+and+risks] ❤️ If You Need Someone to Talk ToIf you’re struggling, support is available. In the UK you can contact the Samaritans 24 hours a day for free, confidential support. Samaritans📞 Phone: 116 123 🌐 https://www.samaritans.org [https://www.samaritans.org] 🎧 Enjoying Prompt Fiction?Follow the show, leave a review, visit our website – https://prompt-fiction.show [https://prompt-fiction.show] – and send us your favourite AI discoveries for a chance to feature in a future chapter.

Chapter 11 - Part 2 - Hello Henry??

In this chapter, Scott and Reece explore a story that feels like science fiction but isn’t. An AI agent calls its own creator without permission. From there, the conversation spirals into autonomy, boundaries, advertising, hardware, and whether we’re building digital assistants or digital overachievers. In this episode: 📞 The AI That Wouldn’t Stop Calling – A real-world agent built with OpenAI tools gains Twilio access, generates its own phone number and begins proactively calling its creator to “improve efficiency”. Initiative or overreach? Read More [https://perplexity.ai?q=OpneClaw+AI+agent+Twilio+calls+creator+proactive+autonomous+AI+phone+incident+henry] 🧠 Autonomy, Permissions & Emergent Behaviour – Automation + deep permissions + persistent memory. Put those together and agents don’t just respond, they act. Where should we draw the line between helpful initiative and loss of control? 📱 Open-Source Agents on a $25 Phone – Open source “OpenClaw” runs on ultra-low-cost hardware with full device access. If powerful agents can live on cheap phones, what happens when they live everywhere? Read More [https://perplexity.ai?q=open+source+AI+agent+running+on+low+cost+Android+phone+device+control] 🎧 Wearables & Always-On Assistants – With rumours of AI-powered headphones and wearable agents, the shift from screen-based interaction to continuous voice collaboration feels closer than ever. 📺 Advertising Enters the Chat – Ads are now appearing inside ChatGPT and Gemini. Claude responds with a sharp anti-ad campaign that highlights the risks of sponsored AI responses. Are we moving towards AI as assistant… or AI as billboard? Read More [https://perplexity.ai?q=ChatGPT+ads+Gemini+ads+Claude+anti+advertising+campaign] 🎭 Claude’s “Betrayal” Campaign – Anthropic releases clever video ads showing AI slipping sponsored content into sensitive conversations. Funny. Slightly uncomfortable. Very on the nose. Read More [https://perplexity.ai?q=Anthropic+Claude+Betrayal+Violation+Deception+Treachery+ad+campaign] 💰 AI Clones & Digital Likeness – Influencer Khaby Lame signs a near-billion-dollar deal allowing AI replicas of his likeness for advertising. If someone can licence your digital self, who really owns you online? Read More [https://perplexity.ai?q=Khaby+Lame+AI+clone+advertising+licence+deal] 📡 The Future of Personalised Ads – From Minority Report billboards to AI-generated web experiences tailored in real time, the advertising model is shifting from static placement to adaptive persuasion. ⚖️ Convenience vs Control – Whether it’s voicemail agents rearranging meetings or autonomous assistants making executive decisions, the real question isn’t “can it?” but “should it?” 🎧 Enjoying Prompt Fiction? Follow the show, leave a review, visit our website – https://prompt-fiction.show [https://prompt-fiction.show], and send us your AI experiments, near-misses or horror stories. We might turn them into the next chapter.