Stripe's AI Agents Can Spend Your Money — And Spotify's War on Slop

The briefing opens with Stripe's annual Sessions conference, highlighting two major announcements. The first is Checkout Studio, a no-code visual builder for payment experiences that combines drag-and-drop design, an AI assistant, live transaction replays, and A/B testing. The AI assistant suggests optimizations based on real transaction data, such as reordering payment methods or trimming form fields. The second announcement is the Link Agent Wallet, which allows AI agents to be granted OAuth access to a user's digital wallet to spend money on their behalf. Stripe is also introducing a Machine Payments Protocol for programmatic agent-driven transactions, though user caution around agent spending remains high. Stripe additionally published a public roadmap and an API assessment tool for developers.

Uber held its own annual product event, where CPO Sachin Kansal unveiled 'One Search,' a universal search bar across all Uber services including rides, food, and hotels. Uber also demoed AI voice capabilities for booking rides conversationally, signaling a strategic shift toward positioning itself as a broader travel and logistics platform.

A notable trend this week is the rise of human verification as a product feature. Spotify launched a 'Verified by Spotify' badge for human artists, responding to data from rival Deezer showing that 44% of all newly uploaded music is now AI-generated. Reddit similarly emphasized its 'authentically human' platform identity in its Q1 shareholder letter, doubling down on this positioning as it faces criticism for hosting AI-generated content that ranks highly in answer engine results.

In other product news, Google added file generation to Gemini, allowing users to create and download PDFs, Word docs, and spreadsheets directly in chat. Notion is rumored to be adding computer-use capabilities to its AI agents via Anthropic's infrastructure. Linear launched 'Linear Releases,' connecting CI pipelines to automatically update issue statuses when code reaches production.

Vercel's head of design published a piece on her team's AI workflow, emphasizing the use of multiple models as collaborative reviewers rather than solo agents—for example, pitting Codex and Anthropic against each other. She also noted that design files are becoming outdated as designers write code directly using production as the source of truth.

A benchmark study from Contra Labs found that no single AI model leads across all stages of the product design process: Claude Opus 4.6 tops ideation, Gemini 2.5 Pro leads on mockups, and Claude reclaims the lead on refinement and polish. On pricing trends, 79 of the 500 largest software firms have moved toward usage-based AI pricing, though seat-based pricing still dominates the broader SaaS market at 96.2% of traditional products. Finally, OpenAI is reportedly betting on an ad-supported model, projecting ads to become its largest revenue driver by 2030 at $102 billion, with the internal comparison being Netflix's successful ad-tier launch.