Anthropic launched Claude Managed Agents, a suite of composable APIs for building cloud-hosted agents at scale. Handles sandboxed code execution, state management, credential vaults, and scoped permissions out of the box. Early adopters include Notion, Asana, Rakuten, Sentry, and Vibecode.

Meta released Muse Spark, the first model from Alexandr Wang's Meta Superintelligence Labs. The natively multimodal model supports visual chain-of-thought and multi-agent orchestration. Meta says AI capex will hit $115-135B in 2026, nearly double last year.

Anthropic shipped the Advisor Strategy for the Messages API, pairing Opus 4.6 as an advisor to Sonnet or Haiku. Scores 2.7 percentage points higher on SWE-bench Multilingual than Sonnet alone while costing 11.9% less per task. A practical pattern for production agent deployments.

Anthropic's annualized revenue run rate hit $30B, overtaking OpenAI's $25B for the first time. Over 1,000 enterprise customers now spend more than $1M/year on Claude, doubling from 500 in February. The company's growth rate is unprecedented in enterprise SaaS.

Google shipped Gemma 4 in four sizes with native vision, audio, function calling, 256K context, and 140+ language support. The flagship 31B model scores 80% on LiveCodeBench and 89.2% on AIME 2026. Apache 2.0 license means zero restrictions for commercial use.

MiniMax released M2.7, hitting 56.22% on SWE-Pro and 57.0% on Terminal Bench 2. The model's three-component system (short-term memory, self-feedback, self-optimization) lets it improve over 24-hour windows. Medal rate of 66.6% matches Gemini-3.1, trailing only Claude Opus-4.6 (75.7%) and GPT-5.4 (71.2%).

Anthropic confirmed Claude Mythos, scoring 93.9% on SWE-bench Verified and 97.6% on USAMO. The model autonomously discovered zero-day vulnerabilities in every major OS and browser during testing, including a 27-year-old OpenBSD bug. Access restricted to 50 organizations under Project Glasswing.

• New Anthropic Labs product for prototypes, slides, one-pagers, and mockups.
• Onboarding reads your codebase and design files to auto-build a team design system.
• Powered by Opus 4.7; research preview for Pro/Max/Team/Enterprise.
• Exports to Canva, PDF, PPTX, or standalone HTML.
• HN reaction thread: 375 points, 243 comments.

• Per-token price unchanged ($5/$25 per million) but the new tokenizer produces ~37.4% more tokens per request (483-submission sample at tokens.billchambers.me).
• Anthropic's migration guide cites 1.33× for Claude Code content, 1.45× for CLAUDE.md, 1.47× for technical docs.
• MRCR long-document recall collapsed from 78.3% on 4.6 to 32.2% on 4.7.
• HN: 674 points Apr 17, 606 points follow-up Apr 18.

• $5B fresh investment from Amazon (up to $25B total); Anthropic commits $100B to AWS over 10 years.
• Secures up to 5GW of compute — Trainium2, 3, and 4 chips — for training and running Claude.
• Deal struck at $380B valuation; Anthropic's ARR hit $30B, up 3× from end of 2025.
• Mirrors Amazon's $50B injection into OpenAI in February; two of AI's top three labs now anchored to AWS.