đ” Claude, Google and the Rise of AI "Brain Fry"
Plus: How Uber built its Agentic Design System, Replit Agent 4 pitches to product teams, is Agentic Commerce in trouble?
Hi product people đ,
Considering Microsoft got a head start with Copilot, right now it feels as though it is Anthropic and Google setting the AI workplace agenda. This week saw major releases from both companies - as well as a new release from Replit which is pitching directly to product teams. Weâll take a look at everything you need to know.
But while these releases are impressive, a new Harvard study warns that using too many AI tools at once could be causing what researchers are describing as âBrain Fryâ. Find out what it is, as well as where product managers rank on the list of roles most susceptible to it.
Plus, former Google CEO Eric Schmidt on why he thinks an emerging new trend amongst developers is âmind bogglingâ.
Happy Friday and have a great weekend!
Rich
Watch on YouTube | Follow on Substack Notes
Key reads and resources for product teams
New from the Department of Product Substack this week:
Deep - Homepage UX explored
In 2026, leading SaaS homepages have quietly turned into live product demos, AI showcases, and competitive weaponsâand this Deep Dive dissects over 20 of them to show exactly how. From Linearâs no-signup interactive workflow, to Stripeâs GDP ticker and Notionâs âmeet the night shiftâ AI framing, it breaks down the real patterns behind whatâs working now.
AI Library - Lead Magnet Mini App Generator
Vibe coding is making it easier for product teams to build mini apps as lead magnets. Shopify has an AI name generator tool, Stripe has a VAT Calculator, Mintlify has a GitHub repo to Docs generator. New in the AI library this week is a database of over 80 different examples of these mini apps that you can use as context and inspiration for building your own lead magnet mini apps, along with practical prompts for generating ideas for your own lead magnet mini apps. The full database and prompts to use are included. (Department of Product)
Growth - How Product Teams Scale Global Experiences without âAI Slopâ
Shipping global products is a challenge when traditional translation methods cannot keep pace with your sprint cycles. New industry data reveals that while 97% of organizations prioritize quality, only 57% successfully maintain a consistent brand voice across languages. For Product Managers, this means navigating the risk of low quality AI slop that compromises user trust. Discover how leading tech teams are moving beyond experimental AI to a business critical strategy that integrates directly into your tech stack. This ensures 92% higher linguistic precision and seamless global launches. (Lilt*)
Case study - How Uber built an Agentic System to automate design specs in minutes
Design systems teams are starting to skip weeks of manual spec-writing and go straight from a Figma component link to productionâready documentation. Uberâs uSpec shows how an agentic, local-first AI pipeline can read real component data (tokens, variants, accessibility semantics), interpret it, and then render polished specs directly into Figmaâcovering multiple platforms and screen reader APIs in minutes, without any design data leaving your machine. (Uber Engineering)
Process - How 3 Notion developers Orchestrate their Coding Agents
This demo gives a glimpse at the future of building with agents in Notion. Here Notion creates new tasks from a meeting note, Cursor handles coding, and a Kanban board keeps fast, parallel agent work legible across the team. (YouTube)
Tools you can use - Wondering (DuoLingo for Everything)
Founded by an ex-NotebookLM software engineer, Wondering describes itself as DuoLingo for Everything, letting you turn any topic into a guided path with bite-size visual lessons that can fit into your busy schedule. (Wondering)
UX - Why Agents will be the primary user of all software
In this piece, Box CEO Aaron Levie argues that the real âusersâ of future software wonât be humans, but autonomous AI agents - trillions of them - spinning up compute, signing up for services, moving money, and doing most economically valuable work behind the scenes. The core value is the shift in vantage point: instead of building products, infrastructure, and business models for people, you now have to build for agents as your primary customer - APIâfirst, self-serve, consumption-based, and secure by default.
Design framework - OpenUI
OpenUI is a new framework that lets the AI compose actual interactive UI (charts, forms, tables, dashboards) on the fly, in response to whatever the user asks. Instead of your AI Assistant saying âhereâs a summary of your Q3 metrics,â it builds and renders a real bar chart. The interface adapts to the conversation, not the other way around. (OpenUI)
Analysis - Ex-Googleâs CEO says AI Agents in development is âMind Bogglingâ
New product features and innovation this week
Google has unveiled major upgrades to Workspace allowing it to pull data from multiple apps to create Docs, Sheets, Slides and more.
In Docs, you describe what you want and Gemini drafts it, pulling from your actual files and past work. In Sheets, it builds entire spreadsheets from a plain-language description, and can run complex optimization problems like staff scheduling or budget allocation. Slides gets full deck generation, matching your existing brand style automatically. And Drive itself is being repositioned from a file storage tool into a searchable knowledge base, where you can ask questions and get summarized answers with citations rather than a long list of files to dig through.
In this example, you can describe what you need and Gemini in Sheets can synthesize files and emails to build a formatted spreadsheet with stylised tables and charts:
As part of the release, Google also confirmed that Gemini in Sheets has hit a 70.48% success rate on the benchmark called SpreadsheetBench - a standardized test designed to evaluate how well AI models handle real spreadsheet tasks. This is very impressive considering Microsoft had a headstart with Copilot but with this release it looks as though Google and Anthropic are very much setting the AI workplace agenda.
Claude gets new Excel and Powerpoint and in-context diagrams
Anthropicâs Claude shipped improvements to its Excel and PowerPoint integrations this week, along with interactive charts and diagrams.
In Excel and Powerpoints, context is now shared across files. Previously, if you were working across both Excel and PowerPoint, youâd need to re-explain your data at each step. Now Claude maintains a single continuous conversation across all your open files - so it can read your spreadsheet, build out a financial model, drop the summary into a slide deck, and draft a follow-up email, all without you having to re-explain the dataset each time. Interactive charts and diagrams are almost like mini Artifacts that appear contextually inside a conversation. So in this example you can see some architectural examples brought to life visually alongside a project timeline.
As well as this, Claude Code also gained a handy new ability this week with the addition of the /btw command. The /btw command is a lightweight side-channel for asking quick, one-off questions or having mini-conversations while your main Claude agent keeps working on the primary task (refactoring, implementing a feature, debugging across files etc).
Replit makes it pitch to flatten the product development process with Agent 4
Replit Agent 4 has unveiled an AI coding agent built to keep product teams in creative flow by handling the coordination and execution work that typically slows down development.
Rather than forcing you through a linear, one-step-at-a-time process, it runs multiple sub-agents in parallel - so auth, database setup, and frontend design can all progress at the same time, with results automatically merged and conflicts resolved behind the scenes.
Agent 4 comes with an infinite canvas that lets you explore UI variants and generate multiple design options while the Agent builds in the background. And because the entire project lives in one place, you can produce web apps, mobile apps, slide decks, data apps, and animations all within the same shared context, without fragmenting your work across multiple tools:
For example, PMs can use it to:
Create pitch decks, data dashboards, and mobile apps all in the same project context
Build clickable prototypes without using up engineering resource
Test assumptions with real, deployed apps before committing dev resources
Submit multiple build requests simultaneously and review before anything merge
Gustoâs Principal PM said that so far at least, the concept makes life as PM much easier:
âThis makes my life as a Product Manager 10x easier. Rather than writing requirements and waiting for Figmas, I can show, not tell.â - Gusto PM
For product teams, the pitch is interesting: Replit is positioning it as a way to build clickable prototypes and test assumptions with real deployed apps, without waiting for engineering resource. Whether that holds up at scale is a different question, but the new release reflects the flattening of the product development process.
Perplexity ships a new Personal version of Computer, targeting enterprise customers through Slack
This week, Perplexity hosted its exclusive developer conference, Ask 2026. During the conference, they unveiled two new products: Personal Computer and Computer for Enterprise.
Personal Computer is a local version of its impressive new product, Computer which this time runs locally on a Mac Mini. Computer for Enterprise connects to tools like Snowflake, Salesforce and HubSpot. Teams can interact with it directly through Slack and use it for delegating coding tasks, building dashboards and financial models, and running scheduled workflows asynchronously.
Perplexity says that Computer saved their internal teams $1.6 million in labour costs and performed the equivalent of 3.25 years of work in four weeks. Those are the kinds of numbers you'd want to verify independently, but the direction of travel towards autonomous workers seems to be accelerating with these new releases.
đ Product data and trends to stay informed
Using too many AI tools at once can lead to âAI Brain Fryâ according to a new study from Harvard Business Review and Boston Consultancy Group.
They defined Brain Fry as mental fatigue that results from excessive use of, interaction with, and/or oversight of AI tools beyond oneâs cognitive capacity. The study found that AI can actually reduce burnout (by replacing repetitive tasks) while simultaneously causing brain fry (by demanding constant monitoring). Using more AI tools boosts productivity - but only up to three; after that, the act of juggling multiple tools at once outweighs the benefit.
Only 9% of product managers report experiencing AI brain fry - well below the 14% average and the 26% seen in marketing.
Is agentic commerce in trouble? A judge has blocked Perplexityâs Comet browser from placing orders on Amazon on behalf of a user. A new report by McKinsey on the topic is well worth a read. In it, it finds that European consumers trust AI tools during the shopping journey but not for completing full checkout journeys. Shopifyâs CFO says weâre still âvery, very earlyâ but the agentic payment industry shows no signs of slowing. Stripe is actively hiring for product designers to build out their agentic commerce journeys and with Ramp launching new cards built for agents this week. Watch a demo of it in action.
Andressen Horowitzâs new top 100 generative AI report shows just how far ChatGPT is ahead of its rivals, but Gemini and Claude are catching up.
The consulting industry isnât dying after all. According to the WSJ, Global consulting spend grew 5.5% in 2025, roughly double the growth rate of the previous year. Accenture is a good example of this momentum: in its most recent quarter, the company reported $2.2 billion in new AI-related bookings, up $400 million from the prior quarter. Accentureâs own CEO recently said that using AI is required for a promotion.
Claudeâs daily active users have more than doubled in 2026, rising from ~5 million to ~10 million.
Monthly subscription apps have grown 7x from January 2022 to present and iOS accounts for 77% of launches. ââHealth & Fitness apps have the best yearly lifetime value (LTV) at $35.64, more than 3Ă Gamingâs $11.22. ChatGPT is having an anchoring effect on what users are willing to pay, according to this analysis. The full report on mobile app revenue is worth a read.
90% of Intercomâs pull requests are now authored by Claude Code.
Paid subscribers get the full DoP Substack including: The Knowledge Series for sharpening your tech / AI skills, the AI Prompt and Skills library and DoP Deep dive reports for in-depth analysis to learn lessons from the worldâs top tech companies.
Really interesting and on the nose with Agent 4 from Replit. I am using it for rapid development of functional prototypes and even had it âingestâ our design system. Cut the cycle of discovery to validating ideas and getting work DOR in days, all by myself.