In this issue
Stripe just rolled out a dedicated wallet that lets AI agents securely make purchases on your behalf. While it sounds like a simple payment feature, it crosses a massive threshold for automated workflows.

The gap between generating text and actually spending money has always been a strict boundary in business automation. Letting agents handle the tedious purchasing steps, while you easily approve the final transaction on your phone, unlocks an entirely new layer of operational efficiency.

Topics of the day:

  • Stripe's new secure wallet for AI agents

  • Anthropic embeds Claude directly into Adobe and Blender

  • Google Gemini outputs native business files in chat

  • The AI industry's shift toward hardware and local workflows

  • Curated reads on local AI hardware, DeepSeek v4, Exa for Claude, and more

  • The Shortlist: OpenAI's AWS Bedrock pivot, ElevenMusic, Skymel's LEGATO, and Google Flow Music

Stripe gives AI Agents a wallet

What's happening: Stripe just launched a dedicated wallet through Link for agents that allows your autonomous AI programs to securely make purchases on your behalf. I love that you can approve every transaction on your phone without ever exposing your actual payment credentials to the bots.

In practice:

  • You can integrate this capability directly into your internal tools using a single install command, rapidly unlocking autonomous financial execution for your business.

  • Your custom AI assistants can now use one-time-use cards to handle tedious tasks like booking travel or paying vendors.

  • Every transaction triggers a real-time notification for your final sign off, keeping you completely in control of the budget without the hassle of manual checkout pages.

Bottom line: Most agent demos still stop at "and then a human pays." This closes that loop. If you've been holding off on agent workflows for procurement, travel, or vendor payments, the blocker is gone.

Anthropic embeds Claude inside creative workflows

What's happening: Anthropic just launched native Claude connectors that plug directly into major creative apps like Adobe Creative Cloud and Blender. Instead of copying text from a separate chat window, you can now prompt the AI to orchestrate multi-step tasks right inside your existing software suite.

In practice:

  • You can automate tedious tasks like batch editing headshots or applying background blur across a folder of images without clicking through menus.

  • Marketing teams can rapidly reformat campaign assets by asking the agent to resize and crop a single horizontal video into multiple vertical formats.

  • Technical operators can debug complex scenes and build custom scripts directly within Blender thanks to a new developer fund partnership with Anthropic.

Bottom line: Smartest model on a benchmark stops mattering once your editor or 3D suite has its own. Pick tools by where they live, not by what score they post.

Gemini now generates files directly in chat

What's happening: Google just updated Gemini so it can natively output ready-to-use business files right in your chat window. You can now prompt the AI to create and export documents directly without the usual copy and paste routine.

In practice:

  • Skip the formatting by instantly turning a loose brainstorm into a polished budget proposal as an Excel file or a client summary as a PDF.

  • Automate deliverable creation to bypass the tedious step of manually moving text into your workspace by letting the AI build your final file.

  • Standardize your workflows so you can generate operational guides and immediately distribute them as downloadable Word docs to your entire team.

Bottom line: The annoying part of using Gemini was always the copy-paste tax at the end. Now Gemini hands you the file and the meeting comes off five minutes shorter.

What I read/use this week

CanIRun.ai - A quick check for whether your machine can run a given open-weight AI model locally.

Blueprint.am - An AI hardware design tool for turning prompts into board and chip schematics.

Building AGI-pilled products - Cat Wu, Head of Product at Claude Code, on how to build AI-native products from the ground up.

DeepSeek v4 tops the Vibe Code Benchmark - The new open-weight model takes #1, beating closed peers on coding tasks.

Exa for Claude - A new connector that exposes Exa's index of websites, papers, and people directly to Claude.

The AI hardware and OS race

What's happening: AI labs are realizing cloud compute is too expensive to run chatbots forever, so they are pivoting toward native hardware and deep software integrations. OpenAI is reportedly building an agent-first phone to control the mobile surface, while Anthropic is embedding Claude directly into creative tools to run closer to your local machine.

In practice:

  • You will soon see a push to run AI locally on your own devices because the cost of compute has actually surpassed employee expenses for many tech teams.

  • Embedding models directly into daily tools like Adobe Creative Cloud allows you to automate complex actions across your workspace without leaving the app.

  • Prepping your business data for these native hardware integrations will unlock fresh growth by turning your static software stack into a proactive teammate.

Bottom line: The labs are quietly admitting that pure-cloud chat isn't the endgame. The next year of releases will live on your devices and inside your apps. Worth watching where the new ones choose to land.

The Shortlist

OpenAI pivoted onto AWS Bedrock just a day after dropping its Microsoft exclusivity, letting developers run GPT models directly where their Amazon cloud data already lives instead of constantly moving it around.

ElevenLabs launched ElevenMusic, a new studio that lets you create, remix, and publish songs using fine-tunes of your own sound, while also earning royalties when paid subscribers use your generated tracks.

Skymel released LEGATO, a fast automation tool that takes your business goal, automatically builds a tested agent workflow, and ships the passing version straight to an API for instant use in your products.

Google unveiled Flow Music, a specialized audio generator out of its Labs division that lets you build custom synthesizers and iterate on musical ideas with deep precision, rather than just praying your initial prompt works.

Why this newsletter?

This newsletter is where I (Kwadwo) share products, articles, and links that I find useful and interesting, mostly around AI. I focus on tools and solutions that bring real value to people in everyday jobs, not just tech insiders.

Keep Reading