Anthropic moves into Word
Anthropic launched a beta of Claude for Microsoft Word aimed especially at legal workflows, signaling a push to embed assistant features directly inside office documents (businessinsider.com). Behind the scenes, leaked Claude Code source code suggests Anthropic has unreleased features in development, while some users and an AMD AI director say Claude Code is showing shallower reasoning and higher token use recently — in short, rapid product expansion with mixed power‑user feedback (xda-developers.com) (indiatoday.in).
Anthropic has put Claude inside Microsoft Word in beta, starting with legal work and contract-heavy document review. (businessinsider.com) Business Insider reported the add-in is aimed at lawyers and other document-heavy teams, with Claude working directly inside Word instead of sending users out to a separate chat window. Anthropic’s own site now lists “Claude for Word” alongside Claude for Excel, PowerPoint, Slack, Chrome, and Claude Code. (businessinsider.com) (anthropic.com) The move extends a Microsoft relationship that was already in place. Anthropic said on November 18, 2025 that Claude models became available in Microsoft Foundry and Microsoft 365 Copilot, and the new Word add-in pushes Claude one step deeper into the Office workflow itself. (anthropic.com) For law firms, the pitch is simple: keep drafting, editing, and reviewing in the document where the work already happens. Artificial Lawyer reported on April 11 that Anthropic was explicitly targeting lawyers, a market where tracked edits, clause comparison, and long-form review are routine. (artificiallawyer.com) Anthropic is making that push while expanding Claude across business software and developer tools at the same time. Its homepage now shows a product menu that spans workplace apps, coding tools, and enterprise plans, suggesting the company wants Claude embedded across daily office work, not just in a standalone chatbot. (anthropic.com) At the same time, Anthropic is dealing with scrutiny from developers over Claude Code, its command-line coding assistant. XDA reported that a source map file shipped in the public npm package exposed internal Claude Code source, including unreleased features and feature flags that had not been announced. (xda-developers.com) InfoQ reported the exposed package was version 2.1.88 and said the leak involved about 512,000 lines of TypeScript code after a source map file was included in the npm release. The Hacker News separately reported that Anthropic removed that version and said it was adding measures to stop a repeat. (infoq.com) (thehackernews.com) Some of the loudest criticism has come from power users who say the coding product has regressed in recent weeks. India Today reported on April 12 that users were complaining about shallower reasoning and higher token use, and The Register tied one of the most detailed complaints to Stella Laurenzo, identified through GitHub and LinkedIn as a director of artificial intelligence at Advanced Micro Devices. (indiatoday.in) (theregister.com) A GitHub issue opened by user “stellaraccident” said Claude Code had become “unusable for complex engineering tasks” after February updates. A GitHub profile for that account identifies the user as Stella Laurenzo, and the issue is public in Anthropic’s Claude Code repository. (github.com 1) (github.com 2) The picture now is a company widening Claude’s reach inside the software people already use, while some of its most technical users argue its coding assistant has become less reliable. Anthropic’s next test is whether deeper placement in tools like Word can outpace the questions surfacing around Claude Code. (businessinsider.com) (indiatoday.in)
