Andrew Miles Davis covers a week of releases that somehow topped the previous one, led by OpenAI launching Dot, an always-on assistant that takes action rather than just answering, positioned as a direct response to Meta Muse overtaking ChatGPT as the most downloaded app. He breaks down OpenAI's new $500 a month tier, more than double its previous top plan, framing it as confirmation of a pricing trajectory he has predicted for over a year where the $20 tier eventually offers what free currently does. Other releases include Claude's new Sonnet 5.5 model, which is already performing impressively enough that Andrew plans to test it against his usual Opus setup, Kling 4.0 re-entering the AI video race against Seedance, Eleven Labs releasing version four, Copilot introducing its own coding tool to compete with Codex and Claude Code, Fireflies launching a Whisper-style universal dictation feature, and Gemini's upcoming move to replace Gems with universal skills compatible across models. He also covers Apple raising phone prices partly due to AI-driven chip and memory shortages, Reid Hoffman criticising the tech sector's apocalyptic AI messaging, an Ofcom report showing 21% of UK adults now use tools like ChatGPT as a primary news source despite lower trust than traditional journalism, and Google's publisher payout test revealing that most media outlets are recovering less than 0.1% of lost referral revenue. Subscribe to In AI Nutshell for the AI news that matters to marketers, every Friday in ten minutes.

Andrew Miles Davis opens October by reviewing a genuinely productive September, where a return to full training routine, quiet progress developing the Story Lens into proper intellectual property, and a 21-day daily LinkedIn video run all delivered on what he had hoped for. For October he sets out three priorities, starting with finding the right YouTube format after three attempts, a talking head episode, then two use case videos that were useful but not entertaining, landing on a plan to aim for genuine edutainment rather than pure utility. On LinkedIn he is scaling back from daily posting to a more sustainable weekly rhythm of one FAQ video and one step-by-step tutorial, filling remaining days with curated content and, in a habit he openly admits he has neglected, actually posting about his own keynotes and training sessions. The third focus is pushing agentic AI tools like Muse and Codex properly, setting two or three genuinely significant tasks with a clear end goal rather than spreading himself across too many half-finished experiments, a pattern he is consciously trying to avoid this time. Subscribe to In AI Nutshell for daily ten-minute AI insight from someone building all of this in real time.

Andrew Miles Davis wraps up September with his monthly tool review, led by Meta's Muse, an agentic assistant he had held off testing for a while given his earlier frustration with Open Claw, which he had bought a Mac Mini for specifically and ultimately abandoned due to complexity. Muse's WhatsApp and iMessage style interface removed that friction entirely, and after pushing it with increasingly difficult tasks he found it succeeding roughly 80% of the time, enough that he says he would genuinely pay for it once the current free access ends, provided the same capability isn't matched elsewhere in tools he already pays for. Claude Design returns for a second consecutive month, credited as the driving force behind a 21-day daily LinkedIn video challenge where motion graphics that would previously have required outsourcing now take seconds to produce. Copilot in Excel also holds its place, now underpinning nearly every operational spreadsheet in Andrew's business from podcast tracking to LinkedIn planning to daily habit discipline, run consistently across both his Mac Mini and laptop. Subscribe to In AI Nutshell for a daily ten-minute update on what is actually worth using, from someone running these tools in real client work every day.