Presented by
Good morning, AI enthusiasts. The Pentagon tried to label an American AI company a national security threat. A federal judge just said that's not how it works.
With Anthropic cleared to keep serving federal agencies, the ruling draws a new line on how far the government can go when targeting domestic AI labs.
In today's recap:
Anthropic's court win blocks Pentagon's supply chain ban
Meta's AI model maps human brain responses to sight and sound
Run multiple coding agents with Cline Kanban
Altman's attempt to "save" Anthropic during Pentagon clash
4 new AI tools, prompts, and more
ANTHROPIC
A federal judge blocks Pentagon's Anthropic gambit
Reuters
Recaply: Anthropic just won a court order that blocks the Pentagon from labeling it a supply chain risk, a move that would have cut off all federal agencies from Claude.
Key details:
The ruling came from US District Judge Rita Lin, who stopped the Pentagon's move in a 43-page decision.
That label, if upheld, would have banned every federal agency from using Anthropic products overnight.
Judge Lin called the Pentagon's logic "Orwellian," writing that no law lets an agency brand an American company "a potential adversary."
The order is preliminary, so the case continues, but Anthropic keeps its federal clients for now.
Why it matters: The Pentagon tried to use a supply chain rule to cut Anthropic out of government work. A court said that's not how it works. That matters beyond one company. It sets a new limit on how far agencies can go when targeting domestic AI labs. The AI industry now has a clearer answer on where that line sits.
PRESENTED BY ATTIO
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?
AI RESEARCH
Meta builds a digital twin of the human brain
Recaply: Meta just unveiled TRIBE v2, a model that predicts how the brain responds to what we see, hear, and read, trained on fMRI data from over 700 people.
Key details:
TRIBE v2 (Trimodal Brain Encoder) predicts brain responses across sight, sound, and language at the same time, with the ability to make predictions for new subjects and languages it wasn't trained on.
The model maps 70,000 brain voxels using over 500 hours of fMRI recordings from 700+ people, one of the largest brain-response datasets ever used in AI research.
TRIBE v2 builds on Meta's Algonauts 2025 award-winning architecture and is fully open-source, so researchers can build on it freely.
Meta says the model can compress what used to take years of lab experiments into hours of simulation.
Why it matters: Brain response research used to take years. You need people in MRI machines, hours of data collection, and large sample sizes just to get usable results. Meta is compressing all of that into a model that runs in hours. Whether this leads to practical breakthroughs, in human-computer interfaces or mental health, is still open. But the direction is clear: AI is now doing science about the minds that use it, at a scale no single lab could ever match.
TUTORIAL
Run multiple AI coding agents in parallel with Cline Kanban
Recaply: In this tutorial, you will learn how to use Cline Kanban to run multiple AI coding agents at the same time on separate git branches, without merge conflicts or terminal overload.
Step-by-step:
Install Cline globally by running
npm i -g cline(Node.js 18+ required), then navigate to the root of any git repo in your terminal and runclineto open the kanban board in your browser.Create task cards for each coding task you want to run in parallel, or type a project description in the sidebar chat and ask Cline to break it into tasks automatically.
Hit Play on a task card to start an agent. Cline Kanban creates an isolated git worktree and terminal for each card, so each agent works on its own branch with no risk of conflicts.
Watch the board as agents work. Each card shows the agent's latest action in real time. Click any card to see a live diff of every change made so far.
Leave inline comments directly on the diff to steer an agent mid-task. When it finishes, hit Commit or Open PR to ship the work, then trash the card to clean up the worktree.
Pro tip: Set task dependencies before you start. When a parent task completes, linked child tasks kick off automatically, turning your parallel board into a sequenced pipeline with no manual handoffs.
TOOLS
Trending AI Tools
⚙️ Cline Kanban - Free, open-source local web app that runs multiple CLI agents in parallel
🔊 Voxtral - MistralAI's open-source TTS model
🎤 Transcribe - Cohere's enterprise speech transcription service
🤖 Gemini 3.1 Flash Live - Google's real-time streaming AI
NEWS
What Matters in AI Right Now?
OpenAI CEO Sam Altman revealed that he told employees he was trying to "save" Anthropic during its Pentagon negotiations, even as OpenAI secured the Defense Department contract Anthropic had just lost.
Karpathy argued that vibe coding hasn't killed software engineering but revealed what the job actually is, with the real work being infrastructure, auth flows, DNS, and deployments that AI still can't navigate.
Google is rolling out a feature letting users import chat history and preferences from other AI apps directly into Gemini, making it easier to switch without losing context built elsewhere.
Wikipedia voted 40-2 to ban AI-generated text from article content, while still allowing editors to use AI for copyediting after a human review step.
Mistral released Voxtral TTS, an open-source text-to-speech model supporting 9 languages, with a time-to-first-audio of 90ms and the ability to clone a custom voice from less than 5 seconds of audio.
OpenAI is indefinitely pausing its ChatGPT erotic mode plans, the latest side project shelved as the company shifts focus toward business users and developers.
OpenAI is backing Isara, a San Francisco startup founded by two 23-year-old researchers that builds software for AI agents to communicate and coordinate on complex tasks.
🧡 Enjoyed this issue?
🤝 Recommend our newsletter or leave a feedback.
How'd you like today's newsletter?
Cheers, Jason
