· 001 · blog · 4 min read
Claude Design Launches as Figma-Killer · GPT-Rosalind Enters Drug Discovery · Claude Opus 4.7 Routines Enable Autonomous Workflows
Claude Design Launches as Figma-Killer · GPT-Rosalind Enters Drug Discovery · Claude Opus 4.7 Routines Enable Autonomous Workflows
Published: 2026-04-25 06:00 (Asia/Shanghai) Coverage: 2026-04-24 18:00 — 2026-04-25 06:00
📰 Top Stories
1. Anthropic Launches Claude Design — Generative Tool That Replaces Canva and Figma
Anthropic unveiled Claude Design, a new generative AI tool capable of building full presentations, websites, brand videos, landing pages, and apps from a single text prompt. In live demos, Claude Design produced a brand video for Uber with motion graphics in under 4 minutes and generated investor-ready pitch decks from scratch. The tool supports multi-format export and is being tested by growth teams who previously relied on expensive design agencies — effectively removing design as a bottleneck for go-to-market execution.
2. Claude Opus 4.7 Introduces Routines — AI That Works While You Sleep
Alongside Claude Design, Anthropic announced Routines, a new functionality for Claude Opus 4.7 that enables scheduled, multi-step task execution on Anthropic’s own servers — completely autonomously. Users can set up automated research pipelines that run overnight and deliver summaries by morning, content production schedules that execute without human trigger, and data monitoring tasks that flag anomalies on a schedule. This positions Anthropic directly against agentic AI players like AutoGPT and LangChain, but with the reliability of a frontier model.
3. OpenAI Codex Major Update Displaces Claude Code in Autonomous Coding
OpenAI shipped a significant Codex update that can open apps on a Mac autonomously, find bugs in existing codebases without being told where to look, and fix and commit those changes automatically — all without developer intervention. This represents agentic software development rather than code completion, enabling non-technical founders to describe product features in plain English and watch Codex build, test, and integrate them end to end.
4. GPT-Rosalind: OpenAI’s Drug Discovery Model Launches for US Pharma Companies
OpenAI introduced GPT-Rosalind, named after Rosalind Franklin, an AI model purpose-built for pharmaceutical drug discovery. Currently available exclusively to US-based pharma companies, the model is designed to accelerate the identification and development of drug candidates — a process that traditionally takes years and costs billions. This represents one of the most consequential AI applications of 2026 in terms of real-world health impact.
5. SpaceX Completes $250B xAI Acquisition — $1.25 Trillion Entity Created
SpaceX finalized its $250 billion acquisition of xAI, creating a vertically integrated entity valued at $1.25 trillion — the largest merger in corporate history. The combined organization now integrates Grok models, Starlink infrastructure, and autonomous vehicle capabilities under one umbrella. Market observers are watching for announcements on how the merged entity will leverage its combined computing infrastructure and data pipelines.
6. Claude Mythos 5 Withheld from Release After Triggering ASL-4 Safety Protocol
Anthropic confirmed Claude Mythos 5 — a 10-trillion-parameter Mixture of Experts model — but announced it will not be released publicly or via standard API after triggering the company’s ASL-4 safety protocol. The model features dedicated expert clusters for cybersecurity, academic research, and complex software engineering. Its cybersecurity capabilities can construct complete multi-stage attack chains, marking the first time a frontier lab has built a completed model deemed too capable to deploy.
7. Google DeepMind’s Gemini 3.1 Pro Leads Multimodal Benchmarks
Gemini 3.1 Pro has emerged as the most capable multimodal model Google has ever shipped, scoring 77.1% on ARC-AGI-2 (novel abstract reasoning) and 94.3% on GPQA Diamond (graduate-level science Q&A) — both highest scores ever reported. It features a 1-million-token context window and processes text, image, audio, and video simultaneously through native multimodal architecture. Google also introduced Gemini 3.1 Flash-Lite, delivering 2.5x faster response times for latency-sensitive deployments.
📊 Trend Watch
| Domain | Hot Topic | Attention |
|---|---|---|
| Generative Design | Claude Design replaces Canva/Figma workflows | ⭐⭐⭐⭐⭐ |
| Autonomous AI | Claude Routines, Codex autonomous coding | ⭐⭐⭐⭐⭐ |
| AI in Healthcare | GPT-Rosalind drug discovery model for US pharma | ⭐⭐⭐⭐ |
| Frontier Models | Claude Mythos 5 withheld, Gemini 3.1 Pro benchmarks | ⭐⭐⭐⭐⭐ |
| AI Consolidation | SpaceX-$250B xAI acquisition, $1.25T entity | ⭐⭐⭐⭐ |
🔮 What to Watch
- Claude Design Enterprise Rollout: Anthropic is expected to announce enterprise licensing and API access for Claude Design — watch for pricing tiers and integration partnerships with existing design toolchains.
- GPT-Rosalind Regulatory Path: As the first AI model deployed specifically for drug discovery, regulatory frameworks from the FDA will set precedent for AI-assisted pharmaceutical development pipelines.
- Agentic Commerce Infrastructure: With Visa’s agent payments platform and autonomous AI agents now executing real-world tasks, the infrastructure for machine-to-machine commerce is rapidly materializing.
Briefing generated: 2026-04-25 06:00 (Asia/Shanghai) Data sources: AI-curated from public technology reports and industry analysis