Andrew Miles Davis covers a genuinely significant week in AI news, leading with a Guardian investigation revealing that brands are secretly using AI-generated influencers to produce fake unboxing videos and customer reviews, paying creators under NDAs to keep the public unaware, against a backdrop where 70% of people currently cannot correctly identify a deepfake. He also covers a landmark moment for the legal industry as Garfield AI, the UK's first approved AI law firm, won its first court trial, recovering £7,000 in unpaid fees for a freelancer with AI handling case preparation while a human barrister managed courtroom advocacy. Other major stories include ByteDance's Seedance 2.5 extending its lead as the top AI video generator with 30-second 4K generation from a single prompt, Eleven Labs releasing a 13-hour AI-narrated audiobook of the Odyssey using a licensed clone of Sir Michael Caine's voice, Getty Images signing a licensing deal with OpenAI that tripled its stock overnight, Claude launching an at-mention feature for Slack that lets the AI complete tasks directly within a channel, and Hootsuite rebuilding its platform around an AI agent that takes action across social media rather than just reporting on it. Subscribe to In AI Nutshell for the AI news that matters to marketers, every Friday in ten minutes.

Andrew Miles Davis returns to his good, bad, and ugly format with three angles he has not covered before in the series. The good focuses on the ability to build bespoke tools and applications through vibe coding without hiring a developer, removing a financial and technical barrier that kept good ideas shelved for years, including in Andrew's own experience. The bad covers shadow AI, the unapproved tools employees use inside organisations that create serious security blind spots, often without the company ever knowing sensitive data has left their control. The ugly lands on AI-accelerated misinformation and its impact on shared truth in society, arguing that the speed at which fake content can now be produced outpaces any ability to fact check it, with consequences already visible in how election results and major events are perceived and trusted. Andrew closes with the idea that AI is not inherently good or bad, it is an amplifier, and the future depends on how wisely it gets used. Subscribe to In AI Nutshell for daily ten-minute episodes that give you the full picture of where AI is heading.

Andrew Miles Davis answers three questions from recent training sessions, including one from a 150-person session with the National Film and Television School, YouTube, and the BBC. He addresses whether AI note takers fabricate information due to a lack of detailed prompts, explaining his own workflow of extracting specific insights from transcripts using trained Claude prompts rather than relying on default summaries. He then tackles a genuinely important question for content creators, whether AI-generated content will all start to sound the same, using his belt system analogy to explain the difference between generic white belt prompting and black belt prompting that incorporates lived experience no AI can replicate. The episode closes with a thoughtful answer to whether children should be taught AI in schools, where Andrew's concern lands less on whether it should happen and more on who is actually equipped to teach it. 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 covers three tools on this week's Cool Tools Tuesday, leading with a long overdue and genuinely improved update to ChatGPT's scheduled tasks feature, which now runs consistently rather than producing the random, off-topic results it was previously known for, and can be connected directly to Gmail, Outlook, or Notion for delivery outside the chat window. He also tests Paraspeech, a new voice dictation tool that functions similarly to Whisper but lacks a free tier and the app-launching command feature found in Typeless, concluding it offers nothing his current tools do not already cover. The episode closes with AISA, a free conversational AI skills assessment that interviews users for twenty minutes before producing a LinkedIn-ready certification, which Andrew explores as both a genuinely useful self-assessment tool and a clever piece of marketing, while questioning how much weight a twenty-minute AI-administered test will carry with employers in its current early stage. Subscribe to In AI Nutshell for three new tools every Tuesday with honest verdicts from someone testing them in real work.

Andrew Miles Davis draws on his own history working at MySpace during its peak to ask a question he has been chewing on in offline conversations for a while: is ChatGPT the new MySpace? He argues that large language models, however impressive, are fundamentally a stopgap technology that teaches people how to talk to AI, while the real shift everyone is actually waiting for is AI agents that take action rather than just give answers. Drawing parallels to MySpace's rapid rise from 2006 and equally rapid decline by 2009, he explores whether chatbots represent the first chapter of a much bigger story rather than the destination itself, and what that means for how businesses and individuals should be thinking about their AI strategy right now. Subscribe to In AI Nutshell for daily ten-minute episodes that connect AI's present to digital history and what comes next.

Andrew Miles Davis covers a week defined by AI being pulled back rather than pushed forward, starting with the US government ordering Anthropic to cut off foreign nationals from its powerful Mythos model just days after its public release, citing national security, and what that might signal about AI becoming segregated along national lines. He covers new consumer trust data showing six in ten people are now put off by seeing the word AI in brand messaging, alongside a report finding 59% of a fresh TikTok account's For You feed is AI generated content, three times higher than YouTube, with categories featuring real people showing the lowest rates. He also breaks down the UN's warning against using pleasantries with AI chatbots due to water consumption, offering a cynical theory about who that messaging really benefits, plus MidJourney's surprising pivot into healthcare body scanning, a coalition of major publishers pushing for AI copyright accountability, and Facebook's new AI search mode pulling answers from real group discussions rather than generic web results. Subscribe to In AI Nutshell for the AI news that matters to marketers, every Friday in ten minutes.

Andrew Miles Davis gives his most direct verdict yet on the five large language models he pays for, covering ChatGPT, Claude, Copilot, Gemini, and Perplexity with an honest breakdown of what each one does well, where each one falls short, and when he actually reaches for it during a working day. He explains why ChatGPT remains his default despite its tendency to agree with him too much, why Claude is the one AI practitioners tend to name as their top choice but still frustrates him with its memory inconsistency, why Gemini's surrounding ecosystem is more impressive than the model itself, how Copilot has won him over this year through deep work with corporate clients, and why Perplexity is his first stop for any question where he needs a trustworthy answer. The episode closes with a conclusion worth sitting with: there is no single best large language model, only the best one for the specific job in front of you. Subscribe to In AI Nutshell for daily ten-minute AI insight from someone using all of these tools in real client work every day.

Andrew Miles Davis answers eleven randomly generated questions in real time, covering everything from the AI claims that make him immediately suspicious to the worst business advice he ever followed, his accidental expertise in song lyrics, and what a normal working day in 2026 would look like to someone waking up from a coma since 1995. He also gives his honest take on what brands do online that makes them look desperate, why he thinks vibe coding will become completely normal within two years, and the one thing he would force every business to learn about AI if he could. The random questions format is one of his most personal episode types and consistently one of the most listened to, because the answers are unscripted and unfiltered. Subscribe to In AI Nutshell for one of these every month alongside daily ten-minute AI insight built for marketers and content creators.