Samwise Tech/AI/Robotics Newsletter
Tuesday, April 21, 2026
Amazon Invests $5B in Anthropic as AI Startup Pledges $100B in AWS Cloud Spending
Amazon has committed an additional $5 billion to Anthropic in a deal that also requires the AI company to spend $100 billion on Amazon Web Services cloud infrastructure over the coming years. The arrangement deepens a partnership that began in 2023 and positions AWS as Anthropic’s primary compute provider for training and deploying its Claude models. Anthropic joins a growing list of AI labs locked into exclusive cloud relationships: Microsoft hosts OpenAI on Azure, while Google Cloud backs its own Gemini models. The deal gives Amazon a structural advantage in the AI cloud market as demand for GPU clusters continues to grow.
Sources: TechCrunch
ChatGPT Launches Images 2.0 With Improved Text Rendering and Web Search Integration
OpenAI has released Images 2.0, an upgraded image generation model built into ChatGPT that substantially improves the system’s ability to render accurate text within generated images. The model also integrates web search, allowing it to reference current information when composing images tied to recent events or specific real-world subjects. Images 2.0 is available to ChatGPT Plus and Pro subscribers. Text-in-image accuracy has been a long-standing weakness for AI image generators; reviewers noted that the new model handles logos, signage, and body copy more reliably than prior iterations. OpenAI said the model uses reasoning steps to plan complex compositions before generating output.
Sources: TechCrunch
AI Startup NeoCognition Raises $40M Seed to Build Agents That Learn Continuously
AI research startup NeoCognition has closed a $40 million seed round to develop AI agents capable of learning continuously from new experiences rather than relying solely on fixed training data. The company says its architecture allows agents to accumulate knowledge over time without catastrophic forgetting — a common failure mode where neural networks lose previously learned information when trained on new data. The funding will be used to scale compute infrastructure and expand the research team. The raise is among the largest seed rounds in AI research this year, reflecting strong investor interest in post-training adaptation techniques for building more capable, longer-horizon AI agents.
Sources: TechCrunch
Humanoid Robot Completes Beijing Half-Marathon, Setting World Record for Bipedal Machines
A humanoid robot completed a full 21-kilometre half-marathon course during an officially sanctioned road race in Beijing on April 19, setting a world record for bipedal robots in a road race event. The robot, developed by a Chinese robotics firm, maintained a consistent pace throughout and finished without requiring human intervention beyond an accompanying support crew. The achievement marks a significant milestone for locomotion research, demonstrating that humanoid systems can sustain dynamic movement over extended distances in uncontrolled outdoor environments. China has made robotics a strategic national priority, and several domestic firms are competing to demonstrate real-world deployment capabilities for humanoid platforms.
Sources: TechCrunch
YouTube Expands AI Deepfake Detection to Cover Celebrities and Public Figures
YouTube has extended its AI-powered likeness detection system to celebrities and other public figures, allowing them to request removal of synthetic media that replicates their appearance or voice without consent. The system, previously limited to music artists, uses facial and voice recognition to identify AI-generated content that closely matches known individuals. Eligible users can submit a removal request, which YouTube will review against its matching algorithms. The expansion responds to a sharp increase in AI-generated deepfake content targeting public figures. Regulators in the US, UK, and European Union have been developing legislation around non-consensual synthetic media, increasing pressure on major platforms to act.
Sources: TechCrunch
Vercel Confirms Customer Data Stolen Via Breach at AI Integration Partner Context AI
Cloud hosting platform Vercel has confirmed a security incident in which customer data was stolen through a breach at third-party integration partner Context AI. Context AI, which provides AI-assisted development tools, was compromised by attackers who then accessed data belonging to Vercel customers who had connected the integration to their accounts. Vercel said it revoked all affected credentials immediately upon detecting the incident and notified impacted customers directly. The company stated that its core hosting infrastructure remained secure. The breach highlights growing supply-chain security risk in developer tooling ecosystems, where third-party integrations often carry broad access to upstream customer data and deployments.
Sources: TechCrunch
Apple Briefly Pulls AI Calorie-Tracking App Cal AI Over Deceptive Subscription Billing
Apple temporarily removed Cal AI, an AI-powered calorie and nutrition tracking application, from the App Store after determining the app violated its guidelines on subscription billing transparency. Cal AI uses computer vision to estimate caloric content from photographs of food and relies on a recurring subscription model. Apple’s enforcement action centred on inadequate disclosure of subscription terms at the point of purchase — a violation the developer resolved by updating the app’s billing screens. The app was reinstated shortly after the fix was submitted. The case follows a pattern of enforcement actions Apple has taken against AI apps in the health and wellness category this year.
Sources: TechCrunch
Tech Pulse
Top Models: Claude Mythos Preview (93.9%) | Claude Opus 4.7 (87.6%) | GPT-5.3 Codex (85.0%)
Market: NVIDIA $4.3T | Alphabet $3.9T | Apple $3.8T
Curated by JD · samwise.agency
Leave a Reply
You must be logged in to post a comment.