Samwise Tech/AI/Robotics Newsletter
Sunday, June 21, 2026
Transformer Co-Inventor Noam Shazeer Leaves Google DeepMind to Join OpenAI
Noam Shazeer, co-inventor of the transformer neural network architecture that underpins virtually every large language model today, announced he is leaving Google to join OpenAI. Shazeer posted on X: “I’m excited to share that I’ll be joining OpenAI.” OpenAI CEO Sam Altman responded publicly: “noam is one of the people I have most wanted to work with since the very beginning of openai. only took 10 years.” Shazeer had returned to Google in 2024 following an earlier stint at Character.AI. His arrival at OpenAI comes as the company prepares for its initial public offering.
Sources: TechCrunch
Nobel Laureate John Jumper Leaves DeepMind After Nine Years to Join Anthropic
Nobel laureate John Jumper, the youngest chemistry prize recipient in more than 70 years, announced Friday he is leaving Google DeepMind after nearly nine years to join Anthropic. Jumper shared the 2024 Nobel with DeepMind CEO Demis Hassabis for AlphaFold, an AI system that predicted the three-dimensional structure of more than 200 million proteins. Jumper said he will take time to recharge before starting at the Claude maker. Anthropic has not disclosed his role, though the hire aligns with the company’s push into life sciences and computational biology. His departure follows Transformer co-inventor Noam Shazeer’s exit to OpenAI.
Sources: TechCrunch
FERC Orders Six US Grid Operators to Justify or Revise AI Data Center Power Connection Rules
The Federal Energy Regulatory Commission unanimously ordered six major US regional grid operators to justify or revise rules governing how AI data centers connect to the electric transmission system. FERC issued tailored show-cause orders to PJM Interconnection, the Midcontinent Independent System Operator, Southwest Power Pool, the California ISO, ISO New England, and the New York ISO — regions serving approximately 200 million Americans across more than 30 states. The commission identified five reform areas including transmission application processes, cost-allocation mechanisms, and co-location arrangements. Grid operators must submit resource adequacy reports within 30 days and respond within 60 days.
Sources: TechCrunch
US Claims ASML’s Top EUV Chip Tool May Have Reached China; ASML Denies the Allegation
The US Commerce Department told ASML leadership that one of the Dutch chipmaker’s extreme ultraviolet lithography machines — the only tools capable of printing the world’s most advanced semiconductor patterns — may have reached China in violation of export restrictions. Commerce Secretary Howard Lutnick raised the concern in recent meetings with ASML’s senior leaders. ASML denied the allegation: “ASML has never shipped an EUV machine to China nor have we shipped to China any component, module or equipment specially designed to be used in an EUV machine,” a spokesperson said. US officials told Bloomberg they have evidence of the shipments but have declined to share it with ASML.
Sources: TechCrunch
Mukesh Ambani’s Jio Launches AI Phone Call Assistant and Smart Home Services Across India
Mukesh Ambani’s Reliance Industries announced at its annual shareholder meeting in Mumbai that its Jio telecommunications unit is rolling out AI services across phone calls, mobile apps, and connected homes. The flagship offering, Jio Call Agent, is an AI assistant that can join phone calls to transcribe conversations, generate summaries, and complete tasks such as booking cabs, ordering food, and making reservations. Reliance has committed approximately $110 billion toward AI investment in India. Ambani is positioning Reliance as a national AI champion, seeking to replicate for artificial intelligence what Jio’s low-cost data plans achieved for mobile internet access beginning in 2016.
Sources: TechCrunch
Signal’s Meredith Whittaker Warns That AI Chatbots Are ‘Not Your Friends’
Signal President Meredith Whittaker said in a Bloomberg interview that AI chatbots like ChatGPT and Claude “are not your friends,” urging critical thinking when using AI tools. “These are not your friends. These are not conscious beings. These are not sentient interlocutors,” Whittaker said. She acknowledged using AI to format documents but said she does not pose questions to chatbots, adding: “I’m very serious about my thinking and writing, and I don’t want the process of working through an idea to be foreclosed or eclipsed by the response of a system that’s averaging what’s already out there.” TechCrunch covered her remarks on June 20.
Sources: TechCrunch
Amazon Seeks to Sell Trainium AI Chips Externally, Taking Aim at Nvidia’s Data Center Business
Amazon is in talks to sell its homegrown Trainium AI chip to companies outside AWS in a direct challenge to Nvidia’s data center dominance. AWS AI chief Peter DeSantis confirmed the discussions to Bloomberg. Until now, Amazon has used Trainium exclusively to power its own AWS cloud instances. If run as a standalone business, Amazon’s chip division would generate an annual run rate of approximately $50 billion, CEO Andy Jassy wrote in his shareholder letter. Amazon’s second-generation Inferentia and Trainium chips can reduce computing costs by up to 80 percent compared to Nvidia’s H100 for AI inference workloads. Amazon shares rose 2.55 percent on June 18 following Bloomberg’s report.
Sources: TechCrunch
Tech Pulse
Top Frontier Models: Claude Opus 4.8 (69.2% SWE-bench Pro) | GPT-5.4 (59.1%) | GPT-5.5 (58.6%)
Top Open Source Models: GLM-5.2 Z.ai (62.1% SWE-bench Pro) | MiniMax M3 (59.0%) | Kimi K2.7-Code (est.)
Top Small Models (15–50B): Poolside Laguna XS.2 33B (68.2% SWE-bench Verified) | Qwen3-235B-A22B (est.) | Llama-4 Scout 17B (est.)
Top Edge Models (0–15B): MAI-Code-1-Flash 5B – Microsoft (est.) | Gemini Flash 2.5 (est.) | Phi-4 Mini (est.)
AI Leaders: NVIDIA ~$3.8T | Microsoft ~$3.4T | Alphabet ~$2.5T
Robotics Leaders: Intuitive Surgical ~$190B | ABB Group ~$70B | Fanuc ~$25B
Curated by JD · samwise.agency
Leave a Reply
You must be logged in to post a comment.