Skip to content
DailyPrompt

your daily dose of claude

01 community

Unauthorised Group Breaks into Anthropic's Restricted Mythos Model via Stolen Contractor Credentials

An unnamed group operating through a private Discord obtained access to Claude Mythos — Anthropic's most powerful, deliberately withheld model — by exploiting credentials from a third-party contractor, gaining entry on the same day the model was publicly announced. The group provided Bloomberg with screenshots and live demonstrations as proof of access. Anthropic confirmed it is investigating but says there is no evidence its own systems were directly compromised.

TechCrunch →
research

Anthropic Launches Monthly Survey to Track AI's Real-Time Workforce Effects

Anthropic launched the Economic Index Survey, a monthly poll of randomly selected Claude users that captures how AI is changing work today and what respondents expect in the near and distant future. The initiative is designed to detect economic shifts — from task automation to hiring changes — before they surface in traditional labour market statistics, which typically lag by months or years.

Anthropic
A
Anthropic
@AnthropicAI

Last month, we published our look into what 81,000 people told us they want from AI. In new research, we've investigated the economic hopes and worries referenced in their responses. Read more: anthropic.com/research/81k-economics

Apr 23 View on X →
research

$ cat story_03.md

New Research: High-AI-Exposure Workers Gain Most Productivity but Fear Displacement Most

Follow-up analysis of responses from 81,000 Claude users finds that workers in roles with high AI exposure — especially software engineers — report the strongest productivity gains but simultaneously express the greatest anxiety about eventual job displacement. For every ten percentage point increase in AI exposure, perceived job threat rises by 1.3 points, and early-career workers are significantly more worried about their futures than senior colleagues.

$ open anthropic-research →
product No. 04

Claude Code /ultrareview Now Deploys a Cloud Fleet of Bug-Hunting Agents

An update to Claude Code's /ultrareview command now dispatches a parallel fleet of agents in the cloud rather than a single local review pass, with each finding independently verified before a ranked severity report arrives in the CLI or Desktop. Reviews on large pull requests typically complete in 5 to 10 minutes and surface an average of 7.5 issues. Pro and Max users receive three complimentary reviews valid through 5 May, with subsequent runs billed at roughly £4 to £16 per review.

Source: allthings.how →
C
Claude
@claudeai

Interactive charts and diagrams are now in Claude Cowork. Available in beta on all paid plans.

Apr 23 View on X →
05
product

Interactive Charts and Diagrams Now Live in Claude Cowork Beta

Anthropic extended its interactive visualisation feature — available in standard chat since March — to Claude Cowork, rolling out in beta across all paid plans. Users can prompt Claude to generate charts, diagrams, timelines, or flowcharts inline; the visuals are built in HTML, making them interactive and responsive to follow-up questions as the conversation develops.

Claude Help Center →
tutorial

Anthropic Publishes Production-Grade MCP Agent Patterns Guide

Anthropic published a new guide on the Claude blog explaining when and how to connect production AI agents to external systems using the Model Context Protocol. The post compares three common approaches — direct API calls, CLI wrappers, and MCP servers — and recommends MCP for cloud-hosted agents that need standardised authentication, tool discovery, and semantic context at scale.

Read on Claude Blog →
Breaking

Simon Willison: Anthropic's Claude Code Pricing Test Has Done Lasting Damage to Trust

Developer and commentator Simon Willison published a detailed post-mortem on Anthropic's undisclosed test removing Claude Code from a subset of Pro plan signups, arguing the lack of any communication causes far more reputational damage than a straightforward price increase would have. He warns the episode risks handing competitive ground to OpenAI's Codex and raises questions about whether educators and teams can safely invest in Claude Code tooling when features can disappear without notice.

simonwillison.net
C
ClaudeDevs
@ClaudeDevs

New in Claude Code: /ultrareview (research preview) runs a fleet of bug-hunting agents in the cloud. Findings land in the CLI or Desktop automatically. Run it before merging critical changes—auth, data migrations, etc. Pro and Max users get 3 free reviews through 5/5.

Apr 22 View on X →
C
ClaudeDevs
@ClaudeDevs

New blog: Building agents that reach production systems with MCP. When should agents use direct APIs vs CLIs vs MCP? Plus patterns for building MCP servers, context-efficient clients and pairing MCP with skills.

Apr 23 View on X →