Samwise Tech/AI/Robotics Newsletter
Thursday, July 9, 2026
OpenAI Launches Full-Duplex GPT-Live-1 Voice Models Optimized for Long Conversations
OpenAI on Wednesday released GPT-Live-1 and GPT-Live-1 mini, full-duplex voice models capable of speaking and listening simultaneously — a key departure from the pipeline-based Advanced Voice Mode they replace. The models support natural interruption, conversations of 30 to 40 minutes, and live translation across most spoken languages. GPT-Live-1 mini becomes the default for all ChatGPT users; paying subscribers gain access to the larger model. When handling complex queries, the new voice interface routes requests to GPT-5.5 for search, reasoning, or agentic tasks. Product lead Atty Eleti noted that more than 150 million people already use voice or dictation features on ChatGPT.
Sources: TechCrunch Share ↗ ✉︎ Email 💬 Text
SpaceXAI Releases Grok 4.5, Claiming Opus-Class Performance at Half the Cost
SpaceXAI launched Grok 4.5 on Wednesday — its first major model release since the company’s IPO — positioning it as a cheaper alternative for coding, research, and knowledge work. Elon Musk described it as “an Opus-class model, but faster, more token-efficient and lower cost,” and the company’s internal assessment places it “roughly comparable to Opus 4.7, but much faster.” SpaceXAI priced the model at $2 per million input tokens and $6 per million output tokens, undercutting Opus 4.7 at $5/$25. The company claims twice the token efficiency of leading frontier rivals. OpenAI’s GPT-5.5, its most powerful model, was separately expected to launch Thursday.
Sources: TechCrunch Share ↗ ✉︎ Email 💬 Text
AI Chip Maker SambaNova Raises $1B at $11B Valuation, Wins JPMorganChase as Inference Partner
AI chip startup SambaNova Systems raised $1 billion at an $11 billion valuation in a Series F first close led by General Atlantic, just five months after closing a $350 million Series E. JPMorganChase selected SambaNova as its inference-infrastructure partner, deploying SN40L and SN50 systems for on-premises AI workloads — a win CEO Rodrigo Liang called a signal to the entire financial sector. SambaNova’s next-generation SN50 chip, due to begin shipping in the second half of 2026 with SoftBank as its first customer, fits trillion-parameter models onto a single rack. Intel, BlackRock, and Qatar’s sovereign wealth fund QIA are among the round’s participants.
Sources: TechCrunch Share ↗ ✉︎ Email 💬 Text
General Intuition Raises $320M to Build the Foundation Model of Physical AI Using Video Game Data
General Intuition, a San Francisco startup led by CEO Pim de Witte, raised $320 million at a $2.3 billion valuation to build what it calls “the foundation model of physical AI.” Rather than collecting millions of hours of robot footage, the company trained its model on video game controller data — button presses, timing, and spatial cues — to develop transferable spatial-temporal reasoning. That foundation model, after fine-tuning on just eight minutes of real-world footage, successfully powered a quadrupedal robot navigating a live office environment. Lead investor Vinod Khosla argues game-derived action data provides richer signals than internet text. De Witte aims to become the GPT-equivalent layer for robotics developers.
Sources: TechCrunch Share ↗ ✉︎ Email 💬 Text
Automate 2026 Sets All-Time Record with 50,000 Registrants and Debut Humanoid Robot Pavilion
Automate 2026, North America’s largest robotics and automation trade show, drew more than 50,000 registrants and 1,230 exhibitors to McCormick Place in Chicago from June 22–25, setting an all-time record. The Association for Advancing Automation (A3) debuted a Humanoid Robot Pavilion sponsored by Nvidia alongside a standalone Humanoid Robot Forum that attracted more than 1,100 participants. The Automate Startup Challenge crowned Mbodi — a platform that lets workers teach industrial robots via natural language and simple demonstrations — as the 2026 winner. Innovation Award recipients included Standard Bots’ Flux AI and Synapticon’s ACTILINK-JD. Automate 2027 moves to Las Vegas in May.
Sources: Robotics & Automation News Share ↗ ✉︎ Email 💬 Text
Academic 'Humanizer' Tool Strips AI Writing Signatures from Research Papers, Dividing Scientists
A University of Minnesota machine-learning researcher, Jie Ding, released a tool on June 20 that instructs AI systems to erase their own writing signatures from academic papers and grant proposals. It removes constructions linked to AI output — such as “not just X, but Y” phrasing and em dashes — and realigns text with the author’s voice. AI detection firm Pangram caught most, but not all, humanized text in initial tests. Scientists are divided: some praise it as a useful editing aid, while others warn it enables undisclosed AI usage in peer review and grant writing. After Nature raised concerns, Ding added a disclosure note to the GitHub page.
Locus Robotics Cold-Storage Upgrade Lets HelloFresh Expand Chilled Fulfillment Fivefold to 500 SKUs
Locus Robotics announced that a cold-storage hardware modification — a heated motor enhancement enabling its Origin AMRs to charge inside refrigerated spaces — allowed HelloFresh’s Factor brand to expand chilled fulfillment from 100 to 500 SKUs. The deployment began with 13 Locus Origin robots in July 2025; their 3-minute-36-second average mission time from order induction to box drop-off led HelloFresh to add 26 more units within three months. EveryPlate, another HelloFresh brand, is scheduled for Locus integration later in 2026. HelloFresh chose Locus over a larger multi-vendor integrated system, citing faster deployment, lower upfront costs, and greater flexibility than fixed conveyor alternatives.
Sources: Robotics & Automation News Share ↗ ✉︎ Email 💬 Text
Tech Pulse
Top Frontier Models: Claude Sonnet 4.6 (79.6% SWE-V) | GPT-5.5 (58.6% SWE-Pro) | Grok 4.5 (new release, ~Opus 4.7 class)
Top Open Source Models: GLM-5.2 (62.1% SWE-Pro) | Qwen3-235B-A22B (leading reasoning) | Llama 4 Maverick
Top Small Models (15–50B): Qwen3-32B | Llama 4 Scout (17B) | Mistral Small 3.1 (24B)
Top Edge Models (0–15B): Laguna XS.2 (44.5% SWE-Pro) | Phi-4-mini (3.8B) | Gemma 3 4B
AI Leaders: NVIDIA ~$4.9T | Microsoft ~$4T | Alphabet ~$3.3T
Robotics Leaders: Intuitive Surgical ~$200B | ABB ~$90B | Fanuc ~$45B
Curated by JD · samwise.agency