Context Window 16
This edition covers a busy week at the IPG Conference and a striking Ethan Mollick chart on AI’s collapsing cost and rising performance, Amazon’s Alexa+ upgrade and what agentic shopping means for book discoverability, an HBR piece on AI agents reshaping consumer behaviour, ElevenLabs’ new ElevenReader Publishing offer for synthetic-voice audiobooks, Claude 3.7 Sonnet’s switchable reasoning mode, the arrival of GPT 4.5, and the UK launch of OpenAI’s Sora video generator.
I’ve been out most of this week at the IPG Conference, speaking to the Publishers Association of South Africa, and delivering training, and it’s felt hard to keep up with the pace of developments. It’s been a huge week for product announcements from Amazon, Claude, ElevenLabs and OpenAI. But before getting into the detail, if there’s one thing to take away this week, it’s this chart posted by Ethan Mollick on AI cost and performance, showing a 99.7% decrease in cost, and at the same time, the latest models performing at PhD level. Amazon announced Alexa+, a significant upgrade to its voice assistant with more sophisticated, agentic AI functionality. As a relatively early adopter of voice technology, I was excited by Alexa’s promise and a little underwhelmed by the reality. This does feel like a significant step for users and particularly for publishers. It’s being rolled out free with Prime membership, which puts it in 100m+ households. Secondly, Alexa is tightly integrated with the book retail experience: I already get regular voice prompts on new books and offers. Adding agentic shopping capabilities into that mix creates a powerful opportunity for publishers—but also a new challenge. If AI agents act as intermediaries, how should publishers rethink discoverability and optimisation? Coincidentally, HBR published a piece this week on how AI agents change the way people shop, and how businesses should respond. For publishers, this piece raises big questions. What metadata will AI agents prioritize when recommending books? How does this reshape search-driven book discovery? The takeaway is that brands (and publishers selling D2C) need to be thinking about product USP, the shift from SEO to AAO (AI Agent Optimisation), and the new sources of data that AI agents will prioritise. Last week I wrote about the ElevenLabs/Spotify partnership. This week the company announced ElevenReader Publishing, a no upfront cost solution for authors and publishers to create ebooks using synthetic voice. The unit economics are cute but anaemic: $1.10 per listener, but only if they listen for at least eleven minutes. Anthropic’s new Claude 3.7 Sonnet model allows users to switch between fast responses and deeper, step-by-step reasoning. This is significant as it allows a single model to handle a range of enquiries, rather than switching between generative and reasoning models (it also points to what the more unified OpenAI product roadmap might deliver). OpenAI’s new GPT 4.5 model is available for Pro and Developer users. Sam Altman describes it as “the first model that feels like talking to a thoughtful person.” Which is, of course, highly subjective. OpenAI also released its Sora video generator to UK users. It’s really simple to use: this trial video took about thirty seconds to generate. It’s also imperfect: watch the movement of the coffee cup in this one. But each of those was generated from a single sentence prompt, with no effort to optimise. I suspect that as we all learn the platform, more sophisticated outputs will quickly emerge.
This was originally published in my email newsletter. To receive weekly updates on how AI is affecting the publishing industry, sign up here.