Andrew Miles Davis covers a week of AI news led by entertainment, starting with Nicolas Cage refusing to sign an AI waiver connected to an Amazon Spider-Noir series. He suggests the stance could mean Cage does not work with the studio again, and asks whether other actors will face the same choice. He also covers the backlash to the Kardashians using AI generated imagery in the opening of their eighth season, with Khloe Kardashian pushing back and saying people should watch the whole sequence before judging it. Andrew uses it to show how using AI can now become a reputational issue, and how people often react to a clip rather than the full story. Other stories include Sony Music asking for more than a quarter of a million AI tracks impersonating artists to be removed, a number that has doubled in six months, and Anthropic updating its usage policy to prohibit sustained unnecessary cruelty towards its AI models. He covers Joseph Gordon-Levitt's concern that AI companies could profit from encouraging emotional relationships with chatbots, McDonald's facing a lawsuit over claims that AI pricing tools coordinated prices between independent restaurants, and Google research in which an AI system interviewed 98 patients before urgent care appointments. Doctors said the summaries helped them prepare in 75% of cases, and the suggested diagnosis matched the final one 90% of the time. Subscribe to In AI Nutshell for the AI news that matters to marketers, every Friday in ten minutes.

Andrew Miles Davis explains why personal AI agents, such as Meta's Muse and Grokbot, are the biggest shift in generative AI right now. He sets out three stages. Chatbots wait for you to ask. Do-it-yourself agents such as ChatGPT work or Claude Code work across your files and apps. Personal agents are proactive and notice things before you ask, such as spotting tomorrow's meeting and pulling together previous conversations and the latest news about the company. He argues the old assumption was that AI gets more useful when you get better at telling it what to do, and the new one may be that it gets more useful when it knows enough about you that you do not have to. He is open about the trade-offs, covering privacy, security and control, plus the lock-in risk of spending two years teaching one agent how you work. His practical advice is to skip the question of which agent to sign up for and ask what you would actually want AI to notice, starting with the things you repeatedly remember, monitor or discover too late, such as an unpaid invoice. He closes by saying the real advantage becomes judgement about what AI should know, watch and act on. Subscribe to In AI Nutshell for daily ten-minute episodes on where AI is really heading.

Andrew Miles Davis resets his prompt hacks series around one idea, that what you add to the end of a prompt now matters more than the prompt itself, because endings give the model the context it needs. He compares this to how hooks replaced headlines in content marketing, and argues prompt endings will become the next thing everyone talks about. He shares three endings with real examples. The first asks for up to five clarifying questions before the final response, which turned a vague request for a fitness plan into one built around his goals, ability, time and equipment. The second tells the AI to stop and ask if any part of the request is ambiguous, which caught that a five day holiday in Italy was far too broad. The third asks for a brief outline and approval before writing a long piece, which let him shape a seven day meal plan before it was built. He ties all three back to his AI sandwich and points listeners to his prompt endings page on his website. Subscribe to In AI Nutshell for prompt hacks twice a month and daily ten-minute AI insight.