đ” The ChatGPT App Store officially launches and Gmail gets a new AI Agent
Plus: Figma says Prototypes are the new PRDs, the State of AI Coding 2025 and free courses from Replit
Hi product people đ,
Coming up in this weekâs product briefing, Google reveals new experimental products, including a Gmail agent. The ChatGPT App Store officially launches - but is this a platform product teams should invest in? Early adopters include Figma, Instacart, Lovable, and others.
Plus, as the year draws to a close, get a snapshot of some of the data points that matter from reports including the State of AI Coding, Twilioâs conversational AI report and SimilarWebâs Generative AI report 2025.
If youâre celebrating the Christmas holidays, I hope you have a wonderful time!
Rich
Watch on YouTube | Follow on Substack Notes
Key reads and resources for product teams
As the year draws to a close, hereâs some of the most popular posts from the Department of Product throughout 2025. Where necessary, these have been refreshed and updated to ensure theyâre just as relevant right now:
New prompt in the AI library - 10x Leverage Tool for Identifying Asymmetric Opportunities for growth
This prompt applies leverage thinking - finding the smallest inputs that create the biggest outputs. You feed it your goal, context, and what youâve already tried, and it analyzes your situation to identify asymmetric opportunities (rare moves where minimal effort yields outsized results). It calculates actual leverage ratios (effort vs. impact), ranks micro-actions by ROI, maps out a minimum viable sequence of 3-5 moves to get you 80% of the way there, and flags hidden multipliers you might be missing. (Department of Product)
UX - Why Instagramâs new navigation changes are a classic example of data-driven UX
Instagramâs head Adam Mosseri said almost all their recent growth comes from Reels, DMs (now Chat), and recommendations. This post breaks down some of the reasons why a change was recently made. (X)
Learn - A collection of free courses mini courses from Replit on building apps and vibe coding
Replit launched a collection of new, free courses it is calling âReplit Universityâ. The collection of courses include AI foundations, Advanced vibe coding and how to use Replit at work. Naturally, it pushes Replit as a tool since itâs published by them but still worth a look if youâre interested in developing AI skillets. (Replit)
Opinion - Prototypes are the new PRDs
A growing number of product managers are finding that the fastest way to clarity is to build. Inside Figma Make, theyâre pressure-testing assumptions early, building momentum, and rallying teams around something tangible. This piece explains how Figma does PRD-driven product development. (Figma Blog)
Strategy - Why feature roadmaps donât work for early stage startups
Feature roadmaps limit exploration and discovery. If the boundaries of exploration are limited to a predefined feature set, you quickly find yourself boxed in, argues Ankush Gupta.
Business books you can read over the Christmas holiday season to get ahead in 2025
A curated collection of books, shortlisted by the Wall Street Journal and Financial Times as business books of the year:
New product features and innovation this week
Google is launching an experimental new AI agent for Gmail called CC. Itâs currently available as part of a beta trial and is designed to help you start your day by giving you a summary of âYour day aheadâ. This can include not only items from your calendar but also from other aspects of your life including upcoming rent payments or RSVPs to personal invites. The agent does this by accessing Gmail, Calendar and Google Drive - as well as the web. You can also email the agent directly inside Gmail to boost your productivity by asking it to help with anything throughout your day.
Itâs an intriguing concept but it feels as though this risks getting lost amongst the other AI features Google / Gmail shipped recently. The name âCCâ also seems potentially confusing given that CC is a core functionality of sending an email, but early feedback from testers is positive.
Google also announced that another experimental product, Opal, is now available in their core Gemini app. If youâve not used Opal before, itâs a tool that lets you build and share AI mini apps using natural language for a bunch of different use cases. For product teams, some of the examples worth checking include:
a business profiler that will conduct analysis on competitors and write up a summary and
a product research mini app that will conduct research and craft a condensed, synthesised report.
The ChatGPT App Store is here - is this the next big platform for product distribution?
Developers can now submit apps to the ChatGPT App Store. After being announced earlier in October, itâs now possible for developers (and product teams) to submit their ChatGPT Apps to the store.
Some of the early Apps that are relevant for product builders include the official Figma App, the Aha! App for product roadmapping and the Lovable App. Lovable published a dedicated page outlining some of the things you can do with their integration including turning conversations into working applications, bringing ideas to life and creating internal tools on demand. But the question remains whether users actually want to work with apps like this; for transactional, conversational use cases like Instacart it instinctively makes sense but for more precise, intricate use cases like building vibe coded apps, it feels as though working with the native application is a better option. Weâll find out in 2026 just how popular these apps turn out to be.
For companies who are currently assessing their product distribution channels, AI App marketplaces are yet another channel that development teams may need to consider in the future. This is a snapshot of all the potential channels a company might consider when distributing their product. Right now, a typical SaaS company might allocate most of its resources and efforts on desktop and mobile - but in the future, this could shift towards allocating more time towards MCP integrations and AI apps as users become less likely to open up the product theyâre using and instead head to tools like ChatGPT or Slack. This example from Mintlifyâs Nick Khami is an example of that.
One platform that continues to grow in popularity is WhatsApp and this week, ElevenLabs launched its conversational agents on the messaging service. ElevenLabs says that this will enable teams to design an agent once and deploy it across multiple different platforms - including WhatsApp. The agent has both voice and chat abilities in WhatsApp which could make it a strong contender for customer service solutions.
Kindleâs new AI Assistant solves a real user problem - but a cautionary tale from a similar feature in Prime Video
If youâve ever started a book and stopped halfway through because you feel paralysed from the fear of forgetting exactly what happened in the chapters youâve already read, Kindleâs new Assistant might help.
This week, Kindleâs product teams shipped a new AI Assistant called âAsk this Bookâ. It lets you get spoiler-free recaps or summaries about the chapters youâve read so far and plot details and character dynamics. The answers are ânon-shareable and non-copyableâ and only available to readers whoâve purchased or rented books. On the face of it, this sounds like an AI Assistant that actually solves a very real problem, rather than a box-checking exercise for AI features. But, letâs hope Amazon has managed to learn from a similar feature it released earlier in the year: Amazon recently had to pull its Prime Video recaps which did something very similar because they were prone to hallucinations.
DoorDash unleashes Zesty - a sign of things to come for building a family of apps as data moats?
The food delivery app DoorDash has released a new app called Zesty that curates restaurants and allows users to discover them using hyper-specific, AI-powered searches like âA low-key dinner in Williamsburg thatâs actually good for introvertsâ. The app is Gen-Z maximalist in its UI and looks like a valuable enough way to discover restaurants in its own right but the ultimate goal is to drive more users to Doordash. With Zesty, it looks like DoorDash is making a defensive land-grab into the pre-transaction discovery layer - the space where consumers decide where to eat before they decide how to get the food. Zestyâs social features could also generate intent signals and taste profiles that are far richer than transaction data alone, making it a valuable asset for personalization in the core DoorDash app.
Some users arenât happy, though, accusing DoorDash of cloning Build your Corner (another restaurant discovery app). But with AI making it easier than ever to build - or clone - mobile apps and recent data showing that the number of apps created has increased for the first time in 8 years, this could mark the start of a broader trend towards building a family of apps as a way to generate data moats used in AI personalization.
Tools you can use
ManyPI - turn any website into an API, with schema, extraction, and sync built in.
Craft - this is the winner of the Mac OS App of the year 2025. A space for notes, tasks, and big ideas.
Monocle - a beautifully designed desktop app that lets you stay focused by focusing on one window at a time.
đ Product data and trends to stay informed
The number of lines per developer has grown from 4450 to 7839 in 2025 and medium-sized teams with 6-15 developers have increased their output from 7005 lines to 13,227 lines. Measuring teams based on the number of lines of code generated isnât a reliable proxy for adding value to users but it likely shows the multiplier effect of AI-augmented engineering teams.
Global internet traffic grew 19% in 2025 with Google remaining the most visited website in the world. Android dominates web traffic with a 65% share and despite new AI entrants throughout 2025, Chrome remains stubbornly stuck at just under 66% of the browser market. The full report from Cloudflare is an interesting read for product teams on the state of the web in 2025.
More than 80% of engineers on Stack Overflow use or want to use AI but only 29% of that group actually trusts AI to do useful work.
Notion is generating half of its revenue from its new AI products, recently passing $600 million in annual recurring revenue.
83% of businesses say that they want to build more AI-powered customer service options. But a word of warning: in the same study, 33% of users say they have sworn at an AI or human conversational agent in frustration and 49% say that when they used an AI agent, it never resolved their query. Twilioâs full report on Conversational AI is worth a read if youâre considering building your own conversational assistants.
Daily active users on the Sora app peaked at 1 million in early November and have since settled around the 750k mark according to the annual 2025 report from Similar Web. The numbers are certainly not huge compared to ChatGPT but given that the app is only available in North America and Japan and only ever peaked at 1 million, it suggests there is a user stickiness in AI generated slop videos. The full report analyses the generative AI landscape with some interesting nuggets:
NotebookLM is a major Google success story with visitors approaching 130 million
Users from AI search tools stay longer, view more pages, and convert better.
The percentage of Google visitors who also visited ChatGPT has grown from 1.6% in 2023 to 14.2 in 2025
Cursor is the fastest growing software vendor of 2025 according to a new report from Brex. As well as the major, well-known companies, the report also contains some lesser known products that might be of interest to product teams.
Paid subscribers get the full DoP Substack including: The Knowledge Series for sharpening your tech / AI skills, the AI Prompt library and DoP Deep dive reports for in-depth analysis to learn lessons from the worldâs top tech companies.