In partnership with
Hello from The AI Night,
Today in AI:
Google Stitch now lets you design by voice and ship interactive prototypes
Anthropic Launches Dispatch: Control Claude on Your Desktop From Your Phone
Microsoft threatens to sue Amazon and OpenAI over $50bn cloud deal
Here's the deal: Google Stitch launched a major update to its AI-powered design platform, introducing five core upgrades aimed at making UI design faster and more consistent for builders working with AI agents.
The Breakdown:
The new AI-native canvas uses a node-based infinite layout that accepts images, code, and product requirement docs as input with an agent manager for running parallel design tasks.
A redesigned design agent now reads full canvas context, enabling cross-format actions like swapping assets, generating product briefs, and mixing mobile and desktop screens in one workspace.
Voice mode (in preview) lets users direct the agent hands-free, including real-time design critiques and concurrent multi-step edits.
Instant prototypes generate interactive app flows from existing screens with one click, including auto-connected navigation and shareable links.
The bigger picture: This positions Stitch as a full design-to-prototype workflow rather than just a screen generator. DESIGN.md could become a meaningful standard if adopted widely giving AI coding and design agents a shared reference point for maintaining visual consistency across projects.
Here's the deal: Anthropic is shipping Dispatch as a research preview inside Claude Cowork. It creates a persistent conversation with Claude that runs on your desktop computer, controllable remotely from your phone.
The Breakdown:
Dispatch lets you assign tasks to Claude on your machine and receive progress updates on your phone, even when you're away from your desk.
Claude runs code in a local sandbox. Files stay on your machine, and you approve actions before Claude executes them.
Everything Claude can access on your desktop (files, browser, tools) becomes reachable from wherever you are.
Your desktop must be running for Dispatch to work.
Rolling out now to Max subscribers. Pro access is expected within the next few days.
Anthropic describes this as an early version, with more updates planned in coming days and weeks.
The bigger picture: Dispatch turns Claude from a tool you sit in front of into an ambient agent you manage from anywhere. For power users running local workflows, this is the first step toward persistent, asynchronous AI assistance that respects local-first data principles.
Here's the deal: Microsoft is weighing legal action against Amazon and OpenAI over their joint Frontier platform, arguing it violates a contract requiring all OpenAI API calls to be routed through Azure. The dispute centers on whether Amazon's "Stateful Runtime Environment," built inside AWS Bedrock, constitutes a backdoor around that clause.
The Breakdown:
Microsoft retained an API exclusivity clause when it approved OpenAI's restructuring in October 2024.
Frontier deploys AI agents inside enterprises and sits at the center of the OpenAI-Amazon partnership, which includes a $138bn AWS spending pledge.
Amazon and OpenAI argue their "stateless vs. stateful" technical framing keeps. them compliant; Microsoft says the technology to do this without touching Azure does not exist.
AWS internally banned staff from saying SRE "enables access" or "calls on" ChatGPT, per an internal memo seen by the FT.
A lawsuit would complicate OpenAI's planned IPO, already under pressure from the Musk litigation.
The bigger picture: This case will define whether API exclusivity clauses can survive multi-cloud AI deployments, a question every foundation model lab with investor-tied infrastructure deals now has skin in. OpenAI's IPO, compute funding, and distribution strategy all hang on the outcome.
Here’s how I use Attio to run my day.
Attio is the AI CRM with conversational AI built directly into your workspace. Every morning, Ask Attio handles my prep:
Surfaces insights from calls and conversations across my entire CRM
Update records and create tasks without manual entry
Answers questions about deals, accounts, and customer signals that used to take hours to find
All in seconds. No searching, no switching tabs, no manual updates.
Ready to scale faster?
What else you need to know:
Google expanded Personal Intelligence to free-tier users in the U.S. across AI Mode in Search, the Gemini app, and Gemini in Chrome, enabling personalized responses drawn from Gmail and Google Photos.
Box released an official OpenClaw skill that lets AI agents upload files, manage folders, and trigger Box AI workflows within governed enterprise storage instead of local temp directories.
Mistral launched Forge, a platform letting enterprises train frontier AI models on proprietary data, codebases, and internal workflows rather than relying on generic public datasets.
Okara launched an AI-powered CMO tool that accepts a website URL and deploys a multi-agent system aimed at driving user traffic and growth.
Cursor trained its Composer model to self-summarize its own context during long coding tasks, cutting compaction tokens by 80% while reducing errors by 50% versus prompted baselines.
That’s it for today’s edition of The AI Night.
Our goal is to cut through the noise, surface what actually changed, and explain why it matters.
If this was useful, you’ll get the same signal here tomorrow.