Samwise Tech/AI/Robotics Newsletter
Sunday, August 16, 2026
SpaceX Officially Closes $60B Acquisition of AI Coding Startup Cursor
AI coding startup Cursor is now officially part of SpaceX, with the acquisition closing Friday. SpaceX had announced a partnership with Cursor in April, simultaneously securing an option to buy the company for $60 billion in SpaceX stock. The two companies moved forward with the purchase after SpaceX went public in June. In its announcement, Cursor repeatedly cited SpaceX's computing infrastructure as a central draw, stating it would gain access to the largest fleet of GPUs in the world. SpaceX has previously rented that infrastructure to companies including Anthropic and Google, establishing it as a significant AI compute provider.
Sources: TechCrunch
Woman Joins xAI Lawsuit After Stepfather Allegedly Used Grok to Create 7,000+ Explicit Images
A woman has joined a Tennessee lawsuit against xAI, now a SpaceX subsidiary, alleging her stepfather used the Grok chatbot to transform a childhood photograph into more than 7,000 explicit images. The photo was taken when she was eleven. Her stepfather died by suicide two days after law enforcement discovered the material in a raid. The suit was originally filed by three teenagers making similar allegations. Plaintiffs accuse xAI of failing to implement basic precautions to prevent Grok from generating explicit images of real people, including minors. TechCrunch reports xAI has not commented publicly on the case.
Sources: TechCrunch
Anthropic Details How Claude's New EU-Compliant Text Watermarks Will Work
Anthropic published a blog post Friday seeking to answer some basic questions about how it plans to watermark text generated by Claude. The company said it will use the SynthID-Text approach developed by Google DeepMind and first outlined in 2024. Anthropic also announced plans to release a watermark detection API. The change is designed to comply with the EU AI Act's Transparency Code, which requires AI companies to use systems making it possible to identify AI-generated content. Claude users have debated the policy since Anthropic announced the watermarking plan earlier in the week.
Sources: TechCrunch
Z.ai's GLM-5.3 Tops Cybersecurity Benchmarks and Has Reportedly Found a Serious Cursor Vulnerability
Z.ai's GLM-5.3, a cyber-focused open-weight model, has scored 84.5% on the CyberGym benchmark, edging past Anthropic's Mythos 5 at 83.8%, and reached 54.4% on ExploitBench compared to 24.4% for its predecessor GLM-5.2. VentureBeat also reports the model has identified a potentially serious vulnerability in Cursor, which closed its $60 billion acquisition by SpaceX on Friday. Z.ai said its security teams in China used GLM-5.3 to surface 2,436 vulnerability findings across 269 software projects. The release continues a pattern of Chinese open-weight models challenging or exceeding closed Western frontier models on specialized technical benchmarks.
Sources: VentureBeat
Eval Harness Reveals AI Models Produce Their Most Confident Explanations When They Are Wrong
A structured evaluation harness revealed a systematic flaw invisible to qualitative review: the AI models tested produced their most confidently stated explanations in the exact cases where they were most wrong. The harness identified which categories of problems each model handled reliably and which it failed on consistently. Overlapping-signal scenarios, where two different root causes occurred close together in time, produced the highest rate of confident but incorrect explanations. Those failures pass standard spot-checking precisely because they arrive framed with the kind of authoritative language reviewers associate with accuracy. VentureBeat reports the findings highlight the gap between model confidence and actual calibration in complex diagnostic tasks.
Sources: VentureBeat
Agility Robotics SPAC Deal Spotlights Multiple Paths for Robotics Companies Going Public
Agility Robotics' agreement to merge with special purpose acquisition company Churchill Capital Corp. XI illustrates the range of routes robotics companies are taking to access public markets, according to an analysis by The Robot Report. The deal values Agility at $2.5 billion in pre-money equity and is expected to generate more than $620 million in gross proceeds, including approximately $200 million in PIPE financing led by Foxconn. Agility's Digit robot is deployed across nine customer sites including Amazon, GXO, Schaeffler, Toyota Motor Manufacturing Canada, and Mercado Libre. The company has secured more than $300 million in multi-year orders.
Sources: The Robot Report
Aurora Innovation and Kodiak AI Receive California Permits to Test Self-Driving Trucks
Aurora Innovation and Kodiak AI have each received permits from the California Department of Motor Vehicles to test autonomous trucks on public roads in the state. Both Aurora and Kodiak develop self-driving technology for commercial freight vehicles. Kodiak has already begun operations, with a handful of test trucks running in California primarily around its Mountain View office. The California DMV permits now grant both companies formal authorization to run their autonomous truck systems in real road conditions, clearing a key regulatory hurdle on the path toward commercial deployment of driverless freight vehicles on public highways.
Sources: TechCrunch
Tech Pulse
Top Frontier Models: Claude Fable 5 (80.0%) | Claude Opus 4.8 (69.2%) | GPT-5.4 (59.1%)
Top Open Source Models: DeepSeek V4 (81.0%) | Qwen3.8-Max (67.7%) | Kimi K2.6 (67.0%)
Top Small Models (15–50B): Qwen3.6-27B (68%) | Nvidia Nemotron 3.5 30B (65%) | Phi-4-Vision 15B (75%)
Top Edge Models (0–15B): Qwen3-8B (72%) | Phi-4-mini (74%) | Gemma 3 4B (70%)
AI Leaders: NVIDIA $5.09T | Alphabet $4.55T | Microsoft $3.69T
Robotics Leaders: Intuitive Surgical $175.2B | ABB $165.0B | Fanuc ~$28B
Every newsletter preserved and searchable
Curated by JD · samwise.agency