Loading...
Loading...
Anthropic’s Claude Code is accelerating a shift toward managed, agentic automation—cloud-hosted agents that run 24/7, integrate with webhooks and connectors, and handle real-world tasks from coding to financial monitoring. Anthropic touts rapid internal adoption, claiming AI now writes most code and automates reporting, while partnerships (notably a $200M Gates Foundation deal) expand healthcare and education use cases. But product and pricing changes—separating programmatic credits from subscriptions and hiking costs for heavy non-interactive use—have raised concerns about vendor lock-in and developer backlash. Competitors and tooling (SDKs, MCP bridges, community skills) are racing to plug gaps and capture displaced demand.
Agentic tooling like Anthropic's Claude Code shifts developer workflows toward supervising autonomous agents, impacting productivity, tooling choices, and operational costs. Tech professionals must evaluate integration, governance, and cost implications as teams and vendors retool around agentic development.
Dossier last updated: 2026-05-14 20:10:54
Anthropic published 'Claude for Legal,' a repository of reference agents, skills, and connectors that automate common in-house, litigation, regulatory, IP, privacy, and academic law workflows for use with Claude Cowork, Claude Code plugins, or the Claude Managed Agents API. The package includes practice-area plugins that ingest a firm’s playbook, named agents (e.g., Vendor Agreement Reviewer, NDA Triager, Renewal Watcher) for end-to-end tasks, scheduled managed-agent cookbooks, and integrations with productivity and legal systems like Slack, Google Drive, DocuSign, Ironclad, and iManage. Outputs are explicitly drafts for attorney review with source attribution and conservative guardrails; Anthropic disclaims legal positions and requires lawyer oversight. This matters because it packages domain-specific LLM tooling and connectors that can accelerate legal operations while stressing professional responsibility and compliance.
Microsoft is canceling most internal licenses for Anthropic’s Claude Code and directing thousands of developers to transition to GitHub Copilot CLI by June 30, the end of its financial year. The Experiences + Devices team — which manages Windows, Microsoft 365, Teams, Outlook and Surface — will wind down Claude Code access, citing a desire to converge on Copilot CLI as a unified command-line agent and tighter integration with GitHub for repos, workflows and security. Sources say cost-cutting tied to the fiscal year and Copilot CLI’s ability to be shaped internally also influenced the move. The change affects designers and non-developer staff who had used Claude Code for prototyping, raising migration and capability gaps to address.
Anthropic has released Claude for Legal, a collection of reference agents, skills, and data connectors tailored to common legal workflows—commercial, privacy, product, corporate, employment, litigation, regulatory, AI governance, IP, and legal education. The toolkit can be installed as Claude Cowork/Codex plugins or deployed via the Claude Managed Agents API, and includes practice-area plugins, managed-agent cookbooks, connectors to productivity and legal systems (Google Drive, DocuSign, Ironclad, Everlaw, CourtListener), and named agents for tasks like Vendor Agreement Reviewer, NDA Triager, Renewal Watcher, and DSAR Responder. Outputs are explicit drafts for attorney review with guardrails (source attribution, jurisdiction flags, privilege checks). The package aims to speed legal review without replacing lawyer responsibility and supports customization via practice profiles and connectors.
马斯克旗下的xAI推出首个编程代理，旨在与Anthropic抗衡
Tom Warren / The Verge : Sources: Microsoft plans to remove most of its Claude Code licenses and push its developers toward GitHub Copilot CLI, after previously pushing Claude Code — Thousands of Microsoft developers will use GitHub Copilot CLI instead — Microsoft first started opening up access to Claude Code in December …
Frontier AI labs are pushing cloud-hosted managed agents—agent harnesses running serverside like Claude Code—which enable 24/7 automation, webhook/email-driven actions, sandboxing and provider-managed security/patching. They’re positioned as a powerful, sticky platform much like AWS Lambda: easy to adopt but hard to migrate away from once workflows, data and integrations live in a vendor’s cloud. The author warns of vendor lock-in despite agent primitives being portable, since exported data and operational assumptions become harder to untangle. Anthropic’s recent pricing change for non-interactive Claude Code (used by most managed agents) — effectively imposing steep new credits and 5–20x cost increases for heavy users — underscores the risk and opens opportunities for competitors like OpenAI.
Anthropic announced a $200 million, four-year partnership with the Bill & Melinda Gates Foundation to fund grants, Claude usage credits, and engineering support for programs in global health, life sciences, education, and economic mobility. The deal funds AI-driven work to accelerate vaccine and therapy development, improve health-system decision-making, and develop connectors, benchmarks, and evaluation frameworks for healthcare intelligence. Projects will include AI-assisted screening for vaccines and therapies targeting polio, HPV, and preeclampsia, integrations with the Institute for Disease Modeling to make forecasts accessible, and co-developed education tools and public goods like benchmarks and datasets. Anthropic says the initiative expands its Beneficial Deployments efforts to deliver AI benefits where markets fall short.
Anthropic announced a $200 million, four-year partnership with the Bill & Melinda Gates Foundation to supply grant funding, Claude usage credits, and technical support across global health, life sciences, education, and economic mobility. The collaboration will build Claude connectors, benchmarks, evaluation frameworks, and public datasets to accelerate vaccine and therapy research, strengthen health-intelligence for governments and frontline workers, and improve disease forecasting via an Institute for Disease Modeling integration. It also targets neglected diseases (polio, HPV, preeclampsia) and co-develops educational tools and public goods for K-12 in the US, sub-Saharan Africa, and India. Anthropic says this expands its Beneficial Deployments work to extend AI where markets fall short.
Anthropic is changing how Claude subscriptions cover programmatic usage: starting June 15, paid Claude plans will receive a separate monthly credit pool dedicated to programmatic calls (Agent SDK, claude -p, Claude Code GitHub Actions) instead of drawing from subscribers' interactive allowances. Tiers get fixed amounts (e.g., Pro $20, Max 5x $100, Max 20x $200, team/enterprise per-seat variations). Once the programmatic pool is exhausted, usage either pauses or continues at full API rates if additional usage credits are enabled. Users and developers see this as a regression because previous subscriptions subsidized programmatic tokens, and past billing errors (such as a $200 mistaken API charge) have already strained trust.
A Reddit user reports that Anthropic’s Claude helped recover access to a Bitcoin wallet worth about $400,000 after 11 years by reconstructing a lost passphrase. The user walked Claude through memory fragments and contextual hints, which the model used to generate candidate passphrases that ultimately matched the wallet’s seed phrase. This matters because it demonstrates AI’s growing practical role in personal security recovery tasks and highlights risks: powerful generative models can both aid legitimate recovery and potentially be misused to reconstruct sensitive credentials. The incident raises questions for crypto security, wallet design, and responsible AI access controls from companies like Anthropic.
A new pull request adds a Managed Agents MCP (Model Context Protocol) server implementation to the anthropic/claude-cookbooks repo, providing tooling for Claude’s Managed Agents sessions. The change introduces a cma-mcp example with environment variables, Dockerfile, .gitignore, README, and server code (stdio and HTTP entry points) plus guidance (CLAUDE.md and skill.md) to map CMA endpoints into tools and register them with zod schemas. It targets running either Claude Desktop (stdio) or claude.ai web (HTTP + Connector), and documents deployment and development commands (bun run stdio/http/typecheck). This matters because it offers a reusable, containerized bridge exposing Managed Agents primitives via MCP, aiding developers integrating Claude agents into workflows and connectors.
有跑 claude 模型蒸馏的大佬吗？
Anthropic CFO Krishna Rao says internal AI systems now write roughly 90% of the company’s code and handle most execution tasks like generating financial reports, shifting employees from doers to supervisors. Anthropic uses Claude and Claude Code to automate coding and routine reporting—monthly reviews are 90–95% automated and some reports that took hours now take 30 minutes. Rao argues increased automation boosted productivity and hiring, with staff focusing on oversight, analysis, and strategy rather than manual execution. Anthropic is deploying AI agent clusters to run multiple projects, and Rao sees combined AI plus high-skill talent as a competitive edge, though broader economic impacts on jobs remain debated.
@cat88tw: 今早信箱跟私訊被客戶與學生灌爆都 claude 昨晚這核爆級變動，講穿了就是六月中之後所有程式化操作 claude code 或 agent sdk 的應用皆需另外計費，不再計入原本月費方案內 ( 這代表如果你平日常用 claude -p
Claude Code in Action，21节课从零到实战。 全英文授课，还有多国语言字幕（德/法/日/韩/西/葡）。 学完还有证书！这羊毛不薅说不过去啊兄弟们！ 👉建议操作： 先访问 Anthropic 官网 https://t.co/ttykmU9CEY 查找 "Academy" 或 "Courses" 入口 找到 "Claude Code in Action" 课程 https://t.co/mzDkVKiDYq
0x0funky/agent-sprite-forge: Agent Skill for generating 2D sprite sheets, transparent PNG frames, and animated GIFs from prompts.
Could a Claude Code routine watch my finances?
CC-Canary: Detect early signs of regressions in Claude Code
Affirm Retooled for Agentic Software Development in One Week | Hacker News Hacker News new | past | comments | ask | show | jobs | submit login Affirm Retooled for Agentic Software Development in One Week ( medium.com/affirmtechnology ) 7 points by brd529 1 hour ago | hide | past | favorite | discuss help Consider applying for YC's Summer 2026 batch! Applications are open till May 4 Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact Search: