Hi product people đ,
Coming up this week, NotebookLM gets some impressive new video capabilities that can transform source materials into decks with audio overviews.
Spotifyâs Chief Product Officer hints at some new voice-enabled AI features, the new Stack Overflow Survey reveals what developers really think about AI and a $1.9 billion startupâs Head of Product says that its new AI agent can handle 70% of a product managerâs job.
This weekâs briefing is supported by the new AI presentation tool Chronicle. Chronicle helps you make visually stunning, designer-grade presentations with AI for free. Itâs like Cursor for presentations and integrates with tools like Notion, Figma, Slack and others. Try it for free here.
Enjoy the rest of your week!
Rich
Watch on YouTube | Follow on Substack Notes
NotebookLM, ChatGPT and the new UX coming to Spotify
First up, NotebookLM is getting a UI refresh and a new Video Overviews mode. The UI refresh adds new iconography to differentiate between different audio types and the Video Overviews will transform source materials into video presentations that include source images and imagery created with generative AI. To be fair, describing these as âvideosâ may be a bit of a stretch since theyâre more akin to presentation slides but still, theyâre a neat addition and could be a helpful way to create new presentations. Googleâs Senior PM has hinted that integrations with video model Veo are on the way soon.
This week, Google has also released a new way to group search results with Web Guide and a new vibe coding mini app builder called Opal.
ChatGPT is getting a new study mode. Itâs designed for students but could equally be useful for professionals who want to learn more about specific topics. Rather than offering up answers to questions immediately, study mode breaks topics down into easily digestible chunks with interactive prompts and knowledge checks through quizzes. Nvidiaâs CEO recently admitted that learning new concepts is one of his most common use cases for AI.
MeanwhileâŠ
Spotifyâs Chief Product Officer has hinted at a new conversational interface for the product. During the companyâs earnings call this week, Gustav Söderström explained that following the launch of voice features earlier in the quarter which allows users to make voice requests to DJ, the company was building up a âunique datasetâ which he described as âvery, very valuableâ.
As a result, we should expect the Spotify consumer experience to become much more interactive. âYou can already write to Spotify, talk to Spotify. Youâre just going to see that expand,â Söderström said. The new feature would also be able to âreasonâ over a userâs learning history. This reference suggests that Spotify could apply reasoning models that can perform multiple steps and ultimately do more than just recommend music.
This strategy of releasing a v1 of an AI feature and then using the new data gathered to build models that can inform new, improved versions of it is clearly something Spotify is doubling down on.
In other newsâŠ
Microsoft has joined the AI browser wars with an update to Edge. Copilot Mode allows users to browse the web while being assisted by AI. Once enabled, Edge users can ask Copilot questions about open tabs and instruct it to perform actions on its behalf. In other words, it works pretty much in the same way as Dia and Perplexity Comet. Satya Nadella says this is just a first iteration and that thereâs âmore to comeâ. With incumbents like Microsoft reaching feature parity so quickly, itâll ramp up the pressure on Dia and Comet to figure out how to differentiate themselves.
Yelp is rolling out a new AI-powered feature that brings to life user reviews. The new feature stitches together video content posted by real reviewers with audio generated by ElevenLabs. Itâs a pretty unique way to bring reviews to life which tech companies could lean on for reviews of their own products.
*sponsored content
Tools you can use
Chronicle - craft stunning decks in minutes. Chronicle pairs expert storytelling with AI to elevate your next investor pitch, team update, or client proposal. It integrates with tools like Notion, Slack, Figma and others - with professional looking results.*
Julius - describes itself as the AI data analyst that works for you. Connect your data, ask questions in plain English, and get insights in seconds. No coding required. This week, it raised $10 million to develop the product further.
Bugbot - Bugbot reviews pull requests and identifies bugs, security issues, and code quality problems.
Frigade - helps users navigate your product. AI agents learn how your product works and use this experience to help other users.
Key reads and resources for product teams
Deep dive - New AI voice features and use cases explored
In this Deep dive, weâll explore real world examples of how companies like Microsoft, Shopify, Perplexity and others are building voice-enabled features into their products. This can include use cases like customer service, UX enhancers, productivity boosters and more. If youâre interested in exploring some of the potential ways voice could be used in your own product, then this deep dive should hopefully help. (Department of Product)
Process - Geminiâs product leader on why the company is no longer a writing first culture
Google Geminiâs product leader says the company is moving from a writing first culture to a building first culture - thanks to vibe coding. (Thread on X)
Marketing - How to get recommended by ChatGPT
Similar to backlinks in SEO, you want your product and content to get amplified by the voices that AI search engines trust. The best starting points are user-generated content and social media sites like Reddit, LinkedIn, YouTube, and Quora; these all rank highly among the top cited domains in AI search. (Substack)
Skills - How to vibe code securely
a16z partner Joel de la Garza sits down with Socket founder and CEO Feross Aboukhadijeh to dive into the intersection of vibe coding and security. (A16z)
Google Deepmind CEO on the future of programming
đ Product data and trends to stay informed
Engineers like and trust AI less in 2025 than they did in 2024. Only 33% of developers trust AI accuracy in 2025, down from 43% in 2024 and 42% in 2023. AI favorability dropped from 77% in 2023 to 72% in 2024 to just 60% this year. Full Stack Overflow Developer Survey 2025.
66% of developers report that âAI solutions that are almost right, but not quiteâ is their top frustration when using AI coding tools. In these cases, the code isnât obviously broken, so developers canât immediately reject it. Instead, they must spend extra time carefully analyzing, debugging, and correcting the AIâs output to make it production-ready. Respondents said that carefully identifying and fixing these issues by hand is often slower than simply rewriting them entirely.
Other nuggets of interest from the report:
Javascript and Node.js remain the top programming languages with 66% and 48%
OpenAIâs GPT models are the most popular AI models with 81.4% of engineers saying they use GPT, followed by Claude at 42.8% and Google Gemini at 35.3%
64% believe AI is not threat to their job, down from 68% last year
Mobile users on ChatGPT are using it an average of 13 days a month - putting it ahead of other platforms like Reddit.
AI is reducing the switching costs in SaaS. AI tools can extract and transfer large amounts of data from one application to another, a process that used to require months of manual work by engineers. Now, AI can do this much faster and with less human effort. One company expects to spend up to 50% less to run new SQL databases in the cloud after using AI tools to port data out of old Microsoft SQL Server databases.
Sales of Metaâs Ray-Ban smart glasses tripled in the first half of the year compared to the same period last year according to a press release from Ray-Banâs parent company. This week, Mark Zuckerberg published a letter outlining his vision of superintelligence. He argues that people who donât wear smart glasses will ultimately be at a âcognitive disadvantageâ to people who do.
FinallyâŠ
A Head of Product at Writer, the $1.9 billion AI agent startup, says that AI can handle at least 70% of a product managerâs work, thanks to its new AI Agentâs advanced ability to browse, code and create presentations.
Unlike most enterprise AI tools that handle basic tasks, Action Agent operates at âlevel four orchestration,â meaning it can coordinate multiple agents, generate its own tools, and execute sophisticated workflows with full visibility and auditability.
If an AI agent can really do 70% of a PMâs work, then this opens up a bunch of possibilities for what the future of a PMâs workload could comprise.
Paid subscribers get the full DoP Substack including: The Knowledge Series for sharpening your tech skills, AI tutorials for putting AI into practice at work and DoP Deep dive reports to learn from the worldâs top tech companies.