Good morning. It’s Thursday, July 9. SpaceXAI and Cursor just shipped their first joint model, and it’s aimed squarely at billable-hour work like legal review and financial analysis, a direct swing at the enterprise ground Anthropic and OpenAI have been building on for months.
Today we are covering:
SpaceXAI and Cursor launch Grok 4.5, their first joint model built for legal, finance, and long-running agentic work
OpenAI releases new full-duplex voice models, replacing ChatGPT’s default voice mode
Block agrees to pay $45 million to settle multistate claims over Cash App security failures
Meta commits over $9 billion to its first data center in Canada, its largest outside the US
Mistral enters robotics with Robostral Navigate, a camera-only navigation model
Let’s dive in
SpaceXAI, Cursor Unveil Grok AI Model for Legal, Finance Tasks
By Carmen Arroyo via Bloomberg
SpaceXAI and Cursor launched Grok 4.5, their first jointly built model, designed for software engineering, legal work, financial analysis, and AI agents rather than general chatbot use.
In legal, Grok 4.5 can review contracts, summarize case files, and analyze regulations, and in finance it handles financial statement analysis, investment research, and risk assessment, priced at $2 per million input tokens and $6 per million output tokens.
The launch comes weeks after SpaceX agreed to acquire Cursor in a deal valuing the startup at $60 billion, positioning xAI to compete more directly with Anthropic and OpenAI on enterprise workloads.
𝕏: Based on strong positive feedback from customers in our beta test program, @SpaceXAI will make Grok 4.5 available to the public tomorrow. It is an Opus-class model, but faster, more token-efficient and lower cost. - Elon Musk (@elonmusk)
OpenAI releases new voice models for more natural live conversations
By Ivan Mehta via TechCrunch
OpenAI released GPT-Live-1 and GPT-Live-1 mini, new full-duplex voice models that can speak and listen at the same time, letting users interrupt naturally and enabling live translation.
GPT-Live-1 mini becomes the new default for ChatGPT’s Advanced Voice Mode, while paid tiers get the larger GPT-Live-1, and both models route complex queries to GPT-5.5 for search, reasoning, or agentic tasks mid-conversation.
More than 150 million people already use ChatGPT’s voice and dictation features, and OpenAI sees voice as a future primary computing interface, though a live Hindi translation demo still sounded stilted and heavily accented.
𝕏: If you have access to GPT Live, test out this prompt. “Hey, I want you to teach me about a concept. <insert concept> but every time I interrupt you, I want you to get progressively more angry at me and tell me to stop talking and let you finish” There is an amazing feeling when an A.I. pushes back. - jason (@jxnlco)
The best way to reach new readers is through word of mouth. If you click THIS LINK in your inbox, it’ll create an easy-to-send pre-written email you can just fire off to some friends.
Cash App owner to pay $45 million to settle allegations of lax security
By Joe Warminsky via The Record
Block, the owner of Cash App, agreed to pay $45 million to settle allegations from regulators in nearly every US state that it misled users about the app’s security and failed to protect them from fraud.
Investigators found Cash App had no working customer support phone line until 2021, pushing users toward scam numbers, and that its “Cash App Friday” promotion let fraudsters target winners despite Block knowing about the abuse.
The settlement requires Block to maintain 24-hour live fraud support and stop overstating Cash App’s safety, and builds on a related CFPB consent order requiring $75 million to $120 million in separate state distributions.
𝕏: Block, the company behind @CashApp, knew its users were being scammed by fraudsters and failed to protect them. Now, my office and a bipartisan group of AGs secured $45 million from Block. I’ll always hold businesses accountable when they prioritize profits over people. - NY AG James (@NewYorkStateAG)
Meta is building its first big data center in Canada amid AI push
By Jonathan Vanian via CNBC
Meta is building its first major data center in Canada, a 1 gigawatt facility in Sturgeon County, Alberta, that will be its largest outside the United States.
The roughly $9 billion project will be powered by a natural gas plant developed by a consortium including Pembina Pipeline, with construction expected to take two to three years.
It’s Meta’s 33rd data center overall, part of its race against Google, Microsoft, and Amazon to build out AI infrastructure, and is projected to support over 3,000 construction jobs at peak.
𝕏: After this and other multi-billion dollar announcements this past week, perhaps a few folks should cool the now debunked claim that a referendum on our province’s future is scaring away investment. It’s not. In fact, the opposite is true. We are leading the rest of the country in attracting investment by a country mile… - Rob Anderson (@FreeAlbertaRob)
Mistral enters robotics with Robostral Navigate, an 8B model that steers robots using just one camera
By Matthias Bastian via The Decoder
Mistral unveiled Robostral Navigate, its first robotics model, an 8B-parameter system that guides robots through unfamiliar environments using only a single RGB camera.
Trained entirely in simulation on about 400,000 recorded paths across 6,000 virtual spaces, the model reaches up to a 79.4% success rate on the R2R-CE navigation benchmark, beating both single-camera rivals and systems using depth sensors or multiple cameras.
Robostral Navigate works across wheeled, legged, and flying robots, and Mistral says reinforcement learning has already added 3.2 percentage points to its success rate with no sign of plateauing, though availability hasn’t been announced.
𝕏: This is an interesting 8b model. Robostral Navigate, for embodied navigation: a robotics navigation model that guides robots to autonomously perform tasks specified with natural language. Single RGB camera. This can run locally on robot with medium class hardware! - Brian Roemmele (@BrianRoemmele)
Trending in AI
Today’s big AI theme: as AI use becomes unavoidable, from campus exams to coding benchmarks to how we measure our own habits, the industry is scrambling to build honest ways to track what AI is actually doing to us.
Brown Professor Suspects Majority of His Class Used AI to Cheat
OpenAI finds roughly 30 percent of popular AI coding test is broken
Khosla-Backed Startup Claims Breakthrough With Largest-Ever AI Model on an iPhone
Thanks for reading to the bottom and soaking in our Newslit Daily fueled with highlights for your morning.
I hope you found it interesting and, needless to say, if you have any questions or feedback let me know by hitting reply.
Take care and see you tomorrow!
How was today’s email?
