In A(i) Nutshell

Andrew Miles Davis covers what he calls the busiest AI news week of the year, opening with an unprecedented wave of simultaneous releases from OpenAI's GPT 6 Sol and GPT 6 Lunar, Anthropic's Opus 5.5, Grok 4.7, a new Gemini model, and a major Meta conference covering AI glasses, Muse, and VR, alongside significant updates from Eleven Labs, Runway, and Midjourney. He covers a new LinkedIn ghostwriter economy emerging as AI makes polished posts cheap to produce, with one study estimating 81% of long LinkedIn posts in July were likely AI generated, and highlights growing scientific debate about whether advanced AI systems could develop something resembling internal experience worth ethical consideration. Other stories include a UK startup raising £20 million to recreate concerts using hyper-realistic digital avatars of artists, Andreessen Horowitz launching a practical AI-focused alternative to traditional college, the first entirely AI-produced British sitcom landing on YouTube to a genuinely split reception, and a significant controversy around Meta's Muse customer service platform after it emerged some responses marketed as AI voice were actually being handled by humans at a third party company. The episode closes with Muse reportedly growing faster on mobile than ChatGPT's early launch period due to its embedded position across WhatsApp, Instagram, and Facebook, its new shopping agent capability working smoothly with Shopify, and Amazon blocking it outright over concerns around credential capture and terms of service violations. Subscribe to In AI Nutshell for the AI news that matters to marketers, every Friday in ten minutes.

Andrew Miles Davis tackles what he considers a genuinely pivotal moment for the golden era of AI, OpenAI's newly published proposal for conversational advertising inside ChatGPT, despite Sam Altman previously calling advertising an aesthetic disaster. He deliberately argues both sides, acknowledging that sponsored suggestions appearing at the exact moment of highest intent, such as a labelled airline offer within a travel itinerary someone can then question in plain English, could genuinely be the most powerful advertising format ever built. He then makes the case against it, walking through the well-worn pattern of Google and Facebook, both of which launched with promises of minimal, transparent advertising before becoming saturated with it once growth slowed and shareholders got involved. Andrew's central concern is a psychological one specific to conversational AI, that people treat a chatbot's answer as a conclusion from an advisor rather than a list of options from a directory, meaning the introduction of paid influence fundamentally changes trust in a way search advertising never did, since users have spent two decades learning to spot ads in a list of links but have no equivalent instinct for spotting influence inside a single confident answer. Subscribe to In AI Nutshell for daily ten-minute episodes tracking exactly how and when the golden era of free, unbiased AI actually ends.

Andrew Miles Davis returns to frequently asked questions after focusing on daily LinkedIn video content for the past few weeks, tackling three questions from recent sessions. He gives a genuinely nuanced answer to whether AI can reduce workplace stress and burnout, arguing that while AI can technically free up hours of time, expectations tend to rise to fill the gap, meaning burnout risk may actually increase unless people deliberately manage their own workload, sharing his own interval-based working method built originally around his RSI that now doubles as a mental wellbeing tool. He clarifies a common misunderstanding about processing speed, explaining that cloud-based AI generation happens on the provider's servers regardless of the user's own device, with local hardware and internet connection affecting perceived speed rather than the actual generation itself. The episode closes with a straightforward recommendation of Eleven Labs as his go-to tool for AI-generated sound effects. Subscribe to In AI Nutshell for daily ten-minute episodes built around the questions real people are asking about AI right now.

Andrew Miles Davis marks 80 weeks of Cool Tools Tuesday with three picks, leading with Meta's Muse, an agentic AI tool built into a WhatsApp-style messaging interface that finally solves a problem he has struggled with for years, connecting AI directly to Apple Notes so scattered notes across folders can be searched, combined, and queried naturally. He covers his first genuinely positive experience with the newly updated Siri, contrasting a frustrating initial test asking for the weather with a much smoother experience retrieving information from Apple Notes the next morning, while noting it still interrupts mid-sentence in a way that lags behind competitors. He also covers a free video generator from There's An AI For That, which produced what he rates as the best result yet on his standard test prompt with sound included and no watermark, and closes with Sherpa by Pocket FM, a fiction-focused collaborative story development tool currently running a $200,000 writing contest, useful for turning a loose creative idea into a structured storyboard though limited strictly to fictional content. Subscribe to In AI Nutshell for three new tools every Tuesday with honest verdicts from someone testing them in real work.

Andrew Miles Davis goes deeper into a story he first covered on the Friday news, three novice hikers stranded overnight on Mount Shasta after relying heavily on Gemini for route planning, gear advice, and nutrition, and finds the full story more revealing than the headline suggests. Gemini reportedly gave them a noon turnaround time, information they ignored entirely, continuing to climb until 7pm before descending in darkness, wandering off route, injuring a knee, losing phone navigation, and running out of water. Andrew rejects both easy conclusions, that AI is useless or that AI had nothing to do with it, arguing the real lesson is about where delegation ends and personal judgement must take over, connecting this directly to his four Ds framework of do, delay, delegate, and delete. He argues that the skill that matters most going forward is not better prompting but knowing precisely what you are prepared to hand over to AI and what decision must always remain yours, particularly when the consequences of being wrong are serious. Subscribe to In AI Nutshell for daily ten-minute episodes that go beyond the headline into what these AI stories actually teach us.

Andrew Miles Davis argues that when using AI to generate names for products, services, or campaigns, most people make the same mistake by asking for more options rather than building a better way to judge the options they already have. He outlines a four-part naming brief covering the product, the audience, the desired outcome, and the point of difference, arguing this alone produces better results than a vague request generating fifty forgettable names. Drawing on a real workshop example where a name loved by the room turned out to mean something offensive in another language, unnoticed by everyone except one person who spoke it, he demonstrates why the next step should be asking AI to attack its own suggestions, surfacing unintended meanings, generic overlap, and confusion risk before any name gets chosen. He connects this to his Story Lens method's exposure stage, arguing that a weak name often reveals an unclear understanding of positioning rather than a naming problem itself, and closes with the central lesson: never ask AI what is best without first establishing best according to what, since criteria must exist before judgement can happen. Subscribe to In AI Nutshell for daily ten-minute AI insight and to the YouTube channel for the full video breakdown with live examples.

Andrew Miles Davis leads this week with a genuinely alarming story from Mount Shasta, where three hikers ignored official safety guidance to turn back by midday after receiving conflicting information from Gemini, resulting in an injury, dead navigation devices, and a night stranded on the mountain waiting for rescue, with the hikers themselves admitting they relied too heavily on AI over their own critical thinking. He also covers OpenAI officially confirming advertising is coming to ChatGPT, a development he has been anticipating and dreading for nearly three years as a signal that the current ad-free era of large language models is ending, drawing a direct parallel to how Facebook's feed evolved once advertising took hold. Other stories include a former TikTok executive's AI photo-coaching app built after her husband's inability to take a decent picture, the official trailer for Artificial starring Andrew Garfield as Sam Altman landing around Christmas, Reddit's research showing AI is making holiday shopping more conversational while community trust still outweighs polished product pages, Adolescence co-writer Jack Thorne raising concerns about screenwriters passing off AI-generated scripts as their own without addressing who trained the underlying models, and a Lancashire company using AI-powered sorting machines to rescue 25 million Lego bricks from landfill for cleaning and resale. Subscribe to In AI Nutshell for the AI news that matters to marketers, every Friday in ten minutes.

Andrew Miles Davis returns to pet peeves for the first time since July with three genuinely fresh frustrations. He starts with training projects and chats in both ChatGPT and Claude, describing hours spent uploading detailed source material only for both models to confirm they can see the information and then ignore it entirely when generating output. He then tests ChatGPT's sycophancy directly, deliberately contradicting himself about a simple fact and watching the model agree with every version regardless of accuracy, contrasted against Claude going too far the other way by rigidly refusing to budge from an outdated goal Andrew had since changed elsewhere. The most costly frustration involves Codex repeatedly failing to fix a teleprompter mirroring issue between his laptop and iPad, making the problem worse with each attempt before running out of credits entirely, a experience Andrew compares to hiring a plumber who breaks something further and then walks off once the paid hours are up. Subscribe to In AI Nutshell for daily ten-minute episodes that tell you honestly what using AI is actually like, frustrations included.

Andrew Miles Davis marks the third anniversary of In AI Nutshell by reflecting on what daily podcasting has actually taught him rather than covering the latest news or tools. He traces the idea back to abandoned attempts at daily video content in 2010 and an early influence from 50 Cent's relentless mixtape output, and shares the surprising discovery that finding something to say every day was never the real challenge, deciding what to focus on was. He weighs the trade-off between audio's consistency and relationship-building strength against video's greater reach, acknowledging he could have grown faster on video but doubts he would have sustained three years of daily output the same way. Looking ahead, Andrew announces a shift in emphasis for the podcast's fourth year, moving beyond tools and prompting mechanics toward judgement, decision-making, and how people navigate ongoing change, tying this directly to his Story Lens method, a framework built around the "stranger comes to town" story structure that he has developed and delivered for over a decade and will be focusing on much more heavily going forward. Subscribe to In AI Nutshell for daily ten-minute AI insight, now heading into its fourth year.

Andrew Miles Davis marks 79 weeks of Cool Tools Tuesday with three specialist tools, starting with ClerkAI, a bookkeeping assistant that prepares accounts for accountants by matching bank transactions to receipts, flagging missing documentation, and drafting communications, all with double opt-in confirmation before touching any login-protected accounts. He covers Music Prompts by Scholarsail, a free tool that builds structured music generation prompts for platforms like Suno and Udio by walking through genre, mood, and energy selections, which Andrew tests with a UK garage prompt and rates as a genuinely solid result for something built entirely from dropdown menus. The episode closes with Magic Tellers, a colouring page and activity generator for kids that turns uploaded photos or text prompts into printable black and white colouring sheets, alongside maze generators, crossword makers, and other activity tools useful for teachers, parents, or anyone working with children. Subscribe to In AI Nutshell for three new tools every Tuesday, some built for daily use and some genuinely niche, but all worth knowing about.