Samwise Tech/AI/Robotics Newsletter
Monday, April 13, 2026
Anthropic Launches Project Glasswing to Put AI’s Power on Defense
Anthropic unveiled Project Glasswing on April 7, a cybersecurity initiative uniting twelve major technology firms — including AWS, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, Microsoft, NVIDIA, and Palo Alto Networks — to deploy AI defensively across critical software infrastructure. The initiative centers on Claude Mythos Preview, an unreleased frontier model scoring 83.1 percent on the CyberGym benchmark versus Opus 4.6’s 66.6 percent. The model has already identified thousands of zero-day vulnerabilities, including a 27-year-old OpenBSD flaw and a 16-year-old FFmpeg bug. Anthropic committed $100 million in model credits and $4 million in open-source security grants, with all partners sharing findings publicly within 90 days.
Sources: Anthropic
Anthropic’s Revenue Hits $30B Run Rate as Google and Broadcom Lock In Multi-Gigawatt TPU Deal
Anthropic has signed a multi-gigawatt agreement with Google and Broadcom for next-generation TPU capacity expected online from 2027, the company announced April 6. The deal extends a November 2025 commitment to invest $50 billion in U.S. AI infrastructure. Anthropic’s run-rate revenue has surpassed $30 billion — up from roughly $9 billion at the end of 2025 — with over 1,000 enterprise customers now spending at least $1 million annually on Claude, a figure that doubled in under two months. The company trains and runs Claude across AWS Trainium, Google Cloud TPUs, and NVIDIA GPUs, with Amazon remaining the primary cloud provider.
Sources: Anthropic
KUKA Unveils Automation 2.0 Strategy, Marrying Physical AI with Industrial Robotics
Industrial robot maker KUKA outlined its Automation 2.0 strategy Monday, centering on a new software platform called KUKA AMP (Automation Management Platform) that integrates artificial intelligence with existing automation systems. The approach shifts from traditional rule-based programming to intent-based automation, where operators specify an outcome and the system determines execution. KUKA recorded its highest-ever R&D investment of €213 million in 2025 and saw China revenue exceed €1 billion for the first time. The announcement follows KUKA’s debut of its AI platform at NVIDIA GTC 2026 and reflects an industry-wide push toward physical AI — adaptive systems that perceive, decide, and act in real-world environments.
Sources: Robotics & Automation News
Generalist AI’s GEN-1 Robot Foundation Model Hits 99% Success Rate in Real-World Tasks
Generalist AI released GEN-1, a foundation model for physical AI that achieves 99 percent task success rates — up from 64 percent for its predecessor — while completing tasks three times faster, the company announced April 11. Trained on 500,000 hours of real-world interaction data sourced primarily from low-cost wearable devices worn by humans rather than expensive teleoperation rigs, the model generalizes across diverse manufacturing tasks with roughly one hour of robot-specific fine-tuning. Demonstrated use cases include kitting auto parts, folding T-shirts 86 consecutive times, servicing robot vacuums over 200 repetitions, and packing phone units more than 100 times in sequence. Early access is now available to selected partners.
Sources: Robotics & Automation News
Waymo Opens Nashville Robotaxi Service as AV Talent War Drives Salaries to $500K
Waymo opened its robotaxi service to the public in Nashville on April 12, marking its eleventh city, while a talent war for autonomous vehicle engineers intensifies with base salaries — excluding equity — reaching between $300,000 and $500,000, according to TechCrunch Mobility. MOIA America and Uber simultaneously launched autonomous microbus testing in Los Angeles ahead of a planned late-2026 robotaxi launch, with human safety operators expected to be removed in 2027. Venture firm Eclipse announced a $1.3 billion physical AI fund split between a $591 million early-stage incubation vehicle and a growth-stage fund. Waymo and Waze also launched a data-sharing pilot routing pothole data from robotaxis to cities.
Sources: TechCrunch
Stanford’s 2026 AI Index: 53% of the World Uses Generative AI as China Closes the Gap
Stanford HAI’s 2026 AI Index, published April 13, reveals generative AI has been adopted by 53 percent of the world’s population — faster than any other technology in history, surpassing the adoption pace of the personal computer, the internet, and the smartphone. China has largely erased the United States’ lead in AI model performance, with Anthropic ranked first globally on Arena, the community-driven LLM evaluation platform, trailed closely by xAI. Environmental costs are mounting: xAI’s training of Grok 4 generated an estimated 72,000 metric tons of CO2, while GPT-4o inference workloads require enough water to sustain 12 million people, according to the report.
Sources: MIT Technology Review
Microsoft Launches Three MAI Foundational Models for Speech, Voice, and Video Generation
Microsoft released three new proprietary AI models April 2, expanding its multimodal stack and positioning itself as a direct competitor to rival AI labs rather than an exclusive OpenAI reseller. MAI-Transcribe-1 transcribes speech across 25 languages at 2.5 times the speed of Microsoft’s Azure Fast offering. MAI-Voice-1 generates 60 seconds of audio in under one second. MAI-Image-2 handles video generation tasks. The trio signals a strategic shift toward first-party model ownership, following the company’s $13 billion investment in OpenAI, and reflects growing enterprise pressure to control proprietary AI infrastructure across the full development and deployment stack.
Sources: TechCrunch
Tech Pulse
Top Models (CyberGym/SWE-Bench): Claude Opus 4.6 (72%) | GPT-4o-2025 (65%) | Gemini 2.5 Pro (63%)
Market: NVIDIA $3.4T | Apple $3.2T | Microsoft $3.0T
Curated by JD · samwise.agency