Anthropic Files Legal Challenge Against Pentagon Supply Chain Designation
This morning Bloomberg confirmed Anthropic has formally initiated a legal challenge against the Department of Defense's supply chain risk designation. CEO Dario Amodei wrote in a blog post titled "Where We Stand with the Department of War" that the company "does not believe this action is legally sound" and sees "no choice but to challenge it in court." Legal analysts at Lawfare wrote that the designation "won't survive first contact with the legal system," pointing to 10 USC 3252's requirement for least-restrictive means — a bar the Pentagon's broad action almost certainly doesn't clear.
Even under the designation, Anthropic says its scope is narrower than initially reported: it applies only to Claude as part of specific Department of War contracts, and does not affect how contractors use Claude to serve other customers. Parallel talks with DoD negotiators are reportedly still ongoing.
1 Million New Users Per Day — Claude Tops Both App Store and Play Store
Anthropic's Mike Kreiger confirmed publicly that more than 1 million people are now signing up for Claude every single day. Free users are up over 60% since January, paid subscribers have more than doubled this year, and daily sign-up records have been broken every day this week. Claude has now claimed the #1 spot on both Apple's App Store and Google's Play Store — the most downloaded free app on both platforms simultaneously.
The surge is directly tied to public reaction to the Pentagon standoff. Users are voting with their downloads — and their wallets.
Claude Launches in PowerPoint — Excel Gets Opus 4.6 Upgrade
Anthropic has shipped Claude in PowerPoint as a research preview for Max, Team, and Enterprise plan users — directly inside the Office add-in. Meanwhile, Claude in Excel has been upgraded to use Opus 4.6 and now supports native Excel operations including pivot table editing and conditional formatting. The ability to switch between Sonnet 4.6 and Opus 4.6 is available within both integrations. This positions Claude as a full-stack Microsoft Office assistant, competing directly with Microsoft's own Copilot.
Compaction API Launches in Beta — Effectively Infinite Conversations
Anthropic has shipped a Compaction API in beta that provides server-side context summarization. When a conversation approaches the token limit, Claude automatically generates a compact summary, replaces the conversation history with a compaction block, and continues seamlessly — giving developers effectively infinite conversation length without manual context management. Enable via the compact-2026-01-12 beta header. Automatic compaction thresholds are configurable per request.
Opus 4.6 Fast Mode — 2.5x Faster Output at Premium Pricing
Anthropic has launched Fast Mode for Claude Opus 4.6 in research preview. The speed parameter in API requests enables up to 2.5x faster output token generation — meaningful for real-time applications, interactive coding sessions, and high-throughput pipelines where speed matters more than cost. Pricing is $30 input / $150 output per million tokens. The 1M token context window beta is also now available for Opus 4.6, matching Sonnet.
Data Residency Controls + Claude Code Analytics API Now Live
Two more enterprise-focused API releases landed this week. Data residency controls let developers specify where model inference runs using the inference_geo parameter — US-only inference is available at 1.1x pricing for models released after February 1, 2026. Separately, the Claude Code Analytics API launched, giving organizations programmatic access to daily aggregated usage metrics including productivity data, tool usage statistics, and cost breakdowns — useful for enterprise billing and ROI reporting.
Anthropic Releases "Effective Context Engineering" Guide — Key Takeaways
Anthropic published a new prompting framework called "effective context engineering" — and several findings run counter to how most people prompt. Using ALL-CAPS and forceful directives like "YOU MUST" or "NEVER EVER" actually overtrigger Claude and produce worse results. Long prompts degrade performance, with reasoning degrading noticeably around 3,000 tokens — the practical sweet spot for instructions is 150–300 words. Structure matters: separate INSTRUCTIONS, CONTEXT, TASK, and OUTPUT FORMAT into distinct sections rather than mixing them in one paragraph. And provide 2–3 concrete examples of desired output instead of explaining what you want in abstract terms.
Claude Code for Personal Life — Viral Health Data Thread
A Medium post in the Data Science Collective titled "Claude Code for Life #1: Managing my Health and Wellness" is making the rounds. The author fed 9+ years of health data into Claude Code and used it to track thyroid disease patterns, synthesize trends across Apple Health exports, and generate chart-based reports. It's a compelling demonstration of Claude Code's agentic capabilities outside of pure software development — and a preview of what the new Claude iOS health data integration enables for non-coders.
The Legal Battle Framing Is Shifting the Whole Industry's Conversation
With Anthropic now in court against the Pentagon, the debate has moved from "will they cave?" to "can the government actually win this?" Lawfare's analysis is being widely shared: the supply chain risk designation under 10 USC 3252 requires least-restrictive means, and a blanket ban likely fails that test. Simultaneously, the fact that Claude was actively used in U.S. military operations in Iran — even while this fight was happening — is raising uncomfortable questions in policy circles about what the ban actually accomplishes versus what it signals.
Claude in Excel & PowerPoint Is a Direct Assault on Microsoft Copilot
The Claude Excel and PowerPoint integrations — now with Opus 4.6 and native operations support — represent a quiet but significant competitive move. Microsoft Copilot's core enterprise value proposition has always been deep Office integration. Anthropic is now offering a credible alternative with a model that routinely outperforms GPT-4 on complex reasoning tasks. The timing, during a week when OpenAI is reeling from its "sloppy" Pentagon deal and Anthropic is attracting record users, could not be better for positioning Claude as the enterprise-safe AI choice.