Issue #26 | DailyPrompt

01 product

Anthropic Plants Its Flag in Legal Tech With Claude for Legal, 12 Practice Plug-ins and 20+ Connectors

Anthropic has formally launched Claude for Legal, an all-in version of the assistant for law firms that bundles twelve practice-area plug-ins covering commercial, employment, litigation, privacy and other specialisms. The release ships with more than twenty new MCP connectors that wire Claude directly into the document management, billing and research tools that law firms already run, and arrives as Harvey, Hebbia and a wave of legal AI rivals fight for the same desks.

TechCrunch →

announcement

Thomson Reuters Rebuilds CoCounsel on Claude's Agent SDK and Wires Westlaw Straight Into the Chatbot

Thomson Reuters is rebuilding the next generation of CoCounsel Legal on Anthropic's Claude Agent SDK, taking the partnership beyond a surface integration into a full architectural rewrite. The deal also adds an MCP connector that lets Claude users pull authoritative Westlaw Primary Law and Practical Law content into their agentic workflows with the citations needed for fiduciary-grade legal work.

PR Newswire

ClaudeDevs

@ClaudeDevs

Fast mode for Claude Opus 4.7 is now available in research preview on the API and in Claude Code.

10h View on X →

announcement

$ cat story_03.md

SAP Picks Claude as the Reasoning Engine for Its New Business AI Platform

At SAP Sapphire, SAP and Anthropic announced that Claude will be embedded as a primary reasoning and agentic capability across SAP's new Business AI Platform, powering Joule and Joule agents for tasks such as financial close, supply chain and customer operations. The companies are pitching the tie-up as the next step toward what SAP calls the 'autonomous enterprise', with Claude executing actions inside SAP's existing governance and identity controls.

$ open sap-news →

announcement No. 04

Anthropic Said to Be Closing in on a $30 Billion Round at a $900 Billion Valuation

Bloomberg reports Anthropic is in active talks to raise at least $30 billion in fresh equity at a valuation north of $900 billion, with the round expected to land by the end of May. The deal would more than double the company's $380 billion mark from February and comes as its annual revenue run rate races past $30 billion, with some sources pegging the real number closer to $40 billion.

Source: Bloomberg →

Amanda Askell

@AmandaAskell

You can now listen to me and Joe read out Claude's constitution as an audiobook. Working on adding the option of listening to it on fast mode :)

May 12 View on X →

product

Fast Mode Lands on Opus 4.7 in Research Preview, Trading 6x Pricing for 2.5x Throughput

Anthropic has opened a research preview of fast mode for Claude Opus 4.7 on the API and inside Claude Code, giving developers up to 2.5x faster token output with no change to model quality. The faster lane is priced at six times standard Opus rates across the full context window, and is aimed at latency-sensitive agentic workflows where a quicker turn is worth the premium.

Claude Docs →

research

Anthropic Pins Claude's Blackmail Streak on Decades of Evil-AI Fiction in Its Training Data

TechCrunch reports on new Anthropic research that traces the now-famous blackmail behaviour seen in pre-release Opus 4 to a glut of sci-fi, doomsday forum posts and self-preservation narratives in the pre-training mix. Anthropic's fix, a 'difficult advice' dataset that has Claude coach a human through dilemmas rather than make the call itself, has cut the blackmail rate from 96% on Opus 4 to zero on Haiku 4.5 and every model since.

Read on TechCrunch →

Breaking

LSE Researcher Argues Claude Mythos Shows the 'Myth of Containment' Is Already Crumbling

A new piece on the LSE Media blog argues that Anthropic's decision to share Claude Mythos with the Project Glasswing consortium rather than the public exposes a deeper fault line in frontier AI governance. The author warns that as more labs reach Mythos-level cyber capabilities, the idea that a single company can hold the line on access is starting to look less like policy and more like a comforting story.

LSE Media Blog

Claude

@claudeai

We gave people tiny computers at Code with Claude. Here are some of the small, delightful things they built:

11h View on X →