🔵 Stripe shows off its new product payment capabilities

Plus: Reddit and Spotify use authentic humans as a differentiator, 2,000 DESIGNmd templates, Deepmind’s Demis Hassabis on agents, AGI and more

Hi product people 👋,

This week saw companies including Stripe and Uber host their annual product showcase events. We’ll take a closer look at some of the announcements that matter to product teams.

Plus, Vercel’s head of design Hannah Hearth shares what products her team uses, Spotify and Reddit double down on human verification as a differentiator to fight back against AI slop and a new study reveals which models excel at different parts of the product design process.

Have a great weekend!

Rich

Watch on YouTube | Follow on Substack Notes

---

Key reads and resources for product teams

Process - The tools that Vercel’s Product Design Team Actually uses
The Vercel design team tried every shiny new AI tool and discovered something surprising: there’s no one right way to design anymore. Hannah Hearth shares what product designers actually use day-to-day, from pairing Claude and Codex for code reviews to running parallel agent threads to stay focused. (Vercel)

Knowledge Series - DESIGN.md Explained: the format reshaping how AI builds UI
Google just open-sourced DESIGN.md - a format that turns design systems into “agent-native” specifications that AI can actually use to generate UI. But the excitement masks real challenges: AI-generated designs look polished even when conceptually flawed, and consistency doesn’t guarantee coherence. I break down what DESIGN.md is, why it matters for product teams, and how it could fundamentally reshape design roles.

New in the AI library

• Analyze how AI agents could disrupt your product’s strategy - use this prompt to think through the strategic implications of the shift from traditional apps to AI agent-based interfaces and what it might mean for your product’s strategy.

• Identify opportunities for Voice Agents in your product - use this prompt to systematically scan your product for places where an AI voice agent would genuinely improve the experience - rather than just bolting voice on because it’s possible. (Department of Product)

Interview - Google Deepmind’s Demis Hassabis on agents, AGI and building products that combine AI with hard science
Demis Hassabis, the Google DeepMind founder who cracked protein folding and won a Nobel Prize, believes we’re only one or two breakthroughs away from AGI. In a wide-ranging conversation, he reveals what’s still missing (continual learning, long-term reasoning, better memory systems), why agents are just getting started, and how to build products that combine AI with hard science. (Y Combinator)

Strategy - The five principles that made Japanese products like the Walkman, Gameboy and Pokemon unexpected hits
Sony’s Walkman shouldn’t have succeeded and colleagues called it ridiculous. Yet it became a global sensation. In this Harvard Business Review piece, Northwestern’s David Schonthal and Tokyo-based writer Matt Alt reveal the five principles behind Japan’s surprising hits - from Game Boy to Pokemon to emoji - showing why convenience beats specs, constraints spark creativity, and intuition matters more than data when predicting the future. (Harvard Business Review)

UX - 10 UI patterns that won’t survive the AI shift
Designer Taras Bakusevych identifies 10 UI patterns under pressure, from setup wizards to notification feeds, and shows what’s replacing them with real product examples. (Substack)

Pricing - Why Headless software is the future and what this means for pricing models
As AI agents become primary software users, traditional business models are breaking. Aaron Levie from Box explores how software pricing must evolve. (X)

Technical showcase - Uber’s Chief Product Officer reveals new capabilities at their annual product event
Uber’s CPO, Sachin Kansal, hosted an event alongside the company’s CEO where they revealed new capabilities including AI voice, the ability to book hotels and a new way to search that the company is calling One Search, which transforms the search bar into a universal finder across all Uber service. (YouTube)

Resource - 2,000 DESIGN.md free files you can use
Search curated references by brand, mood, color, typography, or URL. Open any style for colors, type, spacing, components, and a DESIGN.md your agent can use. (Refero)

---

New product features and innovation this week

Google has released a file generation feature in Gemini that lets users create downloadable documents directly within the chat interface. Instead of drafting content in Gemini, copying it elsewhere, and reformatting, you can now prompt Gemini to generate files in multiple formats - PDFs, Word docs, Excel spreadsheets, Google Workspace files, and others - to download or save to Google Drive immediately.

It’s a little surprising that it took Google so long to do this, but still, it’s a welcome addition to the Workspace Gemini line up.

Stripe shows off its new payment capabilities at Sessions 2026

This week, Stripe held its annual Sessions conference this week - and the event was packed with a bunch of new announcements that are relevant for product teams.

One of the most practically relevant products is Checkout Studio, a visual builder that lets you design, test, and optimize payment checkout experiences without writing code. It combines a drag-and-drop configuration interface with an AI assistant, live transaction replay, A/B testing, and performance analytics.

The tool works by letting you customize checkout fields, payment method selection, and form layout through the Dashboard. The AI assistant suggests optimizations based on your transaction data - things like reordering payment methods to match customer preferences or adjusting required fields. Live transaction replay shows you exactly how real customers experience your checkout, highlighting where they drop off and built-in A/B testing lets you compare different layouts or field configurations against your baseline performance.

Other new features announced include a Link Agent wallet that lets agents spend on your behalf and programmatic transactions using the Machine Payments Protocol. Link functions as a traditional digital wallet - you connect payment methods (cards, bank accounts, crypto wallets, BNPL services) and store checkout details like billing and shipping addresses. When you grant an AI agent access to Link via OAuth authentication, the agent can request permission to spend money. It’s still early days, and data suggests users are still weary about letting agents spend on their behalf, but as trusted companies like Stripe roll out agents with spend limits, that might change pretty quickly.

You can see a thread of examples of what people are building with that here and Stripe has also published a public roadmap of upcoming features here. As well as this, Stripe’s product and engineering teams built their own API assessment tool that lets you check your product’s API docs and flag any potential design issues.

At $2 per check, it’s slightly bizarre but if nothing else, is a neat way for Stripe’s product and marketing teams to flex their technical documentation design skills.

“Authentically human” users is now a product differentiator

Spotify has launched a “Verified by Spotify” badge system to help listeners identify authentic human artists amid rising AI-generated music. The badge appears on artist profiles and search results as a green checkmark with “Verified by Spotify” text. Spotify hasn’t published any data on what percentage of new music is AI generated but a recent report from Deezer revealed that a massive 44% of all music uploaded is now AI generated.

A snippet from Reddit’s latest earnings report

Human verification features are being used as signs of product strength in other products, too. In its latest earnings report this week, Reddit declared its platform as “authentically human”.

Other updates worth knowing

Notion is rumoured to be adding a “Computer” capability to its AI agents, enabling them to control browser-based and desktop tasks within sandboxed environments provided by Anthropic’s infrastructure.

Linear has launched a new feature called “Linear Releases” which connects your CI/CD pipeline to Linear, automatically tracking where each issue stands in deployment. When code changes reach production, issue statuses update automatically - so you see what’s actually live to customers, not just merged code. It does start to feel as though all of the major product development SaaS products are starting to converge in an attempt to own the complete context of how work moves from idea to production.

---

Tools you can use

Adoptly - turn product updates into feature adoption.

Exa for Claude - Give Claude access to billions of websites, docs, papers, people, companies, and more.

Velo - turns your raw screen recordings into watch-worthy videos. Could be useful for things like product demos or new feature explainers.

---

📈 Product data and trends to stay informed

A new study developed by Contra Labs entitled the Human Creativity Benchmark found that no single model excels at all stages of the product design phase. One part of the study looked at landing page results and found that Claude Opus 4.6 leads ideation at 68.9%, winning the open-ended “what should this be?” problem. Gemini 3.1 Pro takes the lead on mockups, hitting 68.9% and Claude reclaims refinement at 60%, edging out competitors on polish and iteration.

New analysis found that enterprise software companies are abandoning flat-fee AI pricing in favor of charging based on actual usage. By the end of 2025, 79 of the 500 largest software firms - including HubSpot, Adobe, and Salesforce - had switched to usage-based models. This number is likely to grow in 2026.

But, outside of the 500 companies tracked in that analysis, seat-based pricing still leads the way. Just 3.8% of traditional SaaS products currently charge on a consumption based model, vs 74% of AI Labs according to Ramp.

Nearly half (46%) of respondents in a new Adobe survey say that they don’t care if a brand uses AI so long as their needs are met.

Meta still burnt through over $4 billion on Reality Labs this quarter, down slightly from $4.2 billion in 2025 according to its latest earnings report. Revenues were also down at $402 million from $412 million in the year prior.

OpenAI is betting on a structural shift away from premium subscriptions toward an ad-supported model that will initially shrink its paying user base but eventually generate far more revenue through advertising.

ChatGPT Go ($8/month) is projected to grow from 7% to 92% of all paying subscribers and ChatGPT Plus ($20/month) would drop from 92% down to just 7%, with Pro making up the rest.

OpenAI expects ads to become its single largest revenue driver, generating $102 billion (36% of total revenue) by 2030. The strategy mirrors Netflix’s 2022 ad-tier launch, which successfully drove both subscriber and total revenue growth, but can they pull it off?

Venture capital is flooding into voice AI startups with over $7 billion invested in Q1 2026 alone. The voice recognition market, currently valued at $22 billion, is expected to nearly triple over the next five years.

---

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 world’s top tech companies.