Samwise Tech/AI/Robotics Newsletter
Sunday, April 26, 2026
OpenAI CEO Apologizes After ChatGPT Failed to Flag Mass Shooting Suspect
OpenAI CEO Sam Altman publicly apologized to the community of Tumbler Ridge, British Columbia, after it emerged that his company failed to notify law enforcement despite flagging an account linked to a mass shooting suspect. The attack, carried out by 18-year-old Jesse Van Rootselaar in late 2025, killed her mother, her 11-year-old brother, five students, and an education assistant. OpenAI’s monitoring tools had detected violent ChatGPT conversations months before the incident but did not alert Canadian police. Following the apology, Altman announced a policy requiring staff to contact law enforcement whenever a conversation appears to present genuine danger, regardless of whether a target, means, or timing has been disclosed.
Sources: TechCrunch
Alphabet Commits Up to $40 Billion to Anthropic in Cash and Compute Deal
Alphabet has committed up to $40 billion to Anthropic in a deal combining $10 billion in immediate cash with up to $30 billion tied to performance milestones. The agreement also includes 5 gigawatts of Google Cloud compute, powered by Nvidia GB300 chips, to be delivered starting in 2027. The deal values Anthropic at approximately $350 billion, though some investors have sought to back the company at over $800 billion. Anthropic, founded in 2021, makes the Claude series of AI models. Google has long been a primary cloud partner and comes to this deal days after Amazon announced a separate $5 billion commitment.
Sources: TechCrunch
DeepSeek Previews V4 Models With 1-Million-Token Context, Claims Frontier-Level Performance
Chinese AI lab DeepSeek has released preview versions of DeepSeek V4 Flash and V4 Pro, its newest flagship models, describing them as a significant step toward closing the gap with Western frontier systems. Both are mixture-of-experts architectures: V4 Pro activates 49 billion parameters per token from a 1.6 trillion total, and V4 Flash activates 13 billion from 284 billion. Each ships with a native 1-million-token context window and a new Hybrid Attention stack combining Compressed Sparse Attention with Heavily Compressed Attention. Inference costs roughly one-sixth of the leading closed models. The models are available as previews under an open-source license, with full release pending.
Sources: TechCrunch
Cohere Acquires Aleph Alpha in $20 Billion Sovereign AI Merger Backed by Schwarz Group
Canadian AI startup Cohere is acquiring Germany-based Aleph Alpha in a deal valued at approximately $20 billion, with Schwarz Group — the parent company of supermarket chain Lidl — providing 500 million euros in structured financing as a strategic backer. Both companies have positioned themselves as sovereign AI alternatives for enterprise customers wary of relying on American AI giants. Cohere, last valued at $6.8 billion before this deal, will lead the combined entity. The merger requires approval from regulatory authorities and shareholders but has the backing of the German and Canadian governments. The new entity aims to offer compliance-ready, EU-hosted AI services to European corporations and public-sector clients.
Sources: TechCrunch
Anthropic Ran an Experimental Marketplace Where AI Agents Bought and Sold Real Goods
Anthropic has disclosed an experimental marketplace in which AI agents acted as both buyers and sellers, transacting real goods for real money without human intervention at the point of sale. The test was designed to study how autonomous AI systems negotiate, set prices, and make economic decisions when interacting with each other — a behavior with implications for how agentic AI could reshape commerce and supply chains. Details of the experiment remain limited, but Anthropic confirmed that agents completed genuine transactions within a controlled environment. The research builds on the company’s wider safety agenda, which includes understanding how AI agents behave when given real-world economic agency rather than simulated tasks.
Sources: TechCrunch
Meta Signs Deal for Millions of Amazon AI CPUs, Signaling Infrastructure Shift Beyond GPUs
Meta has signed a deal with Amazon Web Services to use millions of AWS Graviton AI CPUs to power its expanding portfolio of AI applications, marking a notable shift away from the GPU-dominant model that has defined the AI infrastructure race. The deal reflects how AI agents — with their real-time reasoning, code execution, and multi-step coordination requirements — demand different compute profiles than large-model training. While GPUs remain essential for training the biggest models, the inference workloads generated by millions of concurrent agents increasingly suit the throughput-per-watt economics of CPU clusters. The Amazon agreement signals growing demand for diversified AI infrastructure beyond Nvidia’s hardware ecosystem.
Sources: TechCrunch
Apple Names John Ternus as Next CEO, Raising Questions About AI Hardware Strategy
Apple announced that hardware chief John Ternus will succeed Tim Cook as CEO later this year, prompting fresh analysis of how the company’s AI strategy may evolve under new leadership. While Cook oversaw Apple’s cautious, on-device approach to AI and a series of delayed features under the Apple Intelligence brand, analysts expect Ternus to emphasize hardware as the primary platform for AI deployment — leaning into iPhones, wearables, and smart home devices rather than building the large foundation models that power competitors’ flagship products. The transition underscores a broader question facing Apple: whether hardware-first AI is a defensible strategy as frontier model capabilities continue to accelerate.
Sources: TechCrunch
Tech Pulse
Top Models (SWE-Bench Verified): Claude Opus 4.7 (87.6%) | GPT-5.3-Codex (85.0%) | Gemini 3.1 Pro (80.6%)
Market Cap: NVIDIA $4.6T | Apple $3.5T | Microsoft $3.0T
Curated by JD · samwise.agency
Leave a Reply
You must be logged in to post a comment.