Morning, folks! OpenAI just cut off Cursor after SpaceX bought it, SpaceX is building its own turbine parts because AI data centers can't get enough power, Stripe is giving AI agents a way to spend your money, OpenAI is backing 10 Thai startups; and one AI company is telling new hires to stay away from AI. Let’s get into it.
Today's Top 5
OpenAI Just Cut Off Cursor. The Reason: We Can’t Trust Musk’s Companies. OpenAI is ending its model deal with Cursor after SpaceX bought the coding tool. OpenAI says the issue isn't performance. It doesn't trust SpaceX to honor the deal.
SpaceX Is Building Its Own Turbine Factory to Power AI Data Centers: SpaceX is preparing to make key turbine parts in Texas instead of waiting on suppliers with years-long backlogs. Musk says that could get gas turbines online up to 18 months faster.
Stripe Just Gave AI Agents a Way Into Your Wallet: Stripe is updating Link so AI agents can make purchases without seeing your card details. You approve the agent, set spending controls, and it handles the payment.
OpenAI Just Picked 10 Thai Startups to Build With Its AI: OpenAI is putting 10 Thai startups through an 8-week accelerator, giving them model access, technical support, and help building their products. It's not just giving them access. It's getting closer to the companies' buildings with its models.
An AI Startup Just Banned AI for New Hires: Valon wants most new employees to learn the job before they start using AI. The idea is simple: learn the work well enough to know when the AI gets it wrong.
OpenAI is ending its contract providing models to Cursor, shutting off access November 12, following SpaceX's acquisition of the coding tool. The reason is stated directly, OpenAI says it can't be confident SpaceX will honor its terms, pointing to Twitter breaking a prior contract after Musk's takeover and Musk admitting under oath earlier this year that xAI had distilled OpenAI's own data to train its models.
Nearly four years of partnership ended over trust, not performance. OpenAI gave the maximum contractual notice allowed specifically to soften the landing for developers and said it's the people building on Cursor, not the companies involved, who are the real ones caught in the middle.
SpaceX Is Building Its Own Turbine Factory to Power AI Data Centers
SpaceX is laying the groundwork for a turbine-blade and vane factory in Bastrop, Texas, to deal with a problem AI data centers keep running into: power. Instead of waiting on suppliers with years-long backlogs, SpaceX wants to make some of the key turbine parts itself.
Musk says that could get gas turbines online up to 18 months faster. And the timing makes sense. AI companies can keep buying GPUs, but that doesn't help much if they can't get enough power to run them. The AI infrastructure race is starting to look less like a chip race and more like an energy race.
Stripe Just Gave AI Agents a Way Into Your Wallet
Stripe is updating Link so AI agents can make purchases without getting access to your actual card details. You approve the agent, and Stripe handles the payment through virtual cards or tokens. That means an agent could find a hotel, book it, and pay without ever seeing your card number.
Stripe is also adding controls so you can decide what an agent is allowed to buy and how much it can spend. Agents have been able to search, compare, and recommend things for a while. The awkward part was getting them to actually pay. That's what Stripe is working on now.
OpenAI Just Picked 10 Thai Startups to Build With Its AI
OpenAI is teaming up with Thailand’s Ministry of Higher Education, Science, Research, and Innovation on an 8-week accelerator for 10 local AI startups working across health, wellness, and education. The founders will get access to OpenAI models, technical support, and help turning their ideas into products people can actually use.
What's interesting is that OpenAI isn't just giving these startups access to its models. It's getting closer to the founders building with them. If even a few of these companies take off, OpenAI gets more products, developers, and customers building around its technology in Thailand.
An AI Startup Just Banned AI for New Hires
Valon is making most new hires work without AI until their managers think they know the job well enough to spot when the AI is wrong. The mortgage software company made the change after finding people were using expensive models for basic tasks and, in the process, skipping the work that actually teaches you how the job works. Engineers are exempt because their code already goes through peer review.
The policy sounds strange coming from an AI company, but the reason is pretty simple. Valon's CEO says new hires need to make mistakes, debug problems, and build their own judgment before handing that work to a model. There's a financial upside too: the company expects its annual AI spending to fall from $15–20 million to around $4–5 million.
Other AI Signals:
Andreessen Horowitz raised $1.1 billion for a new AI infrastructure fund, betting that the biggest opportunities won't all sit at the model layer. The firm is targeting the infrastructure underneath AI, from compute and networking to the systems needed to run increasingly demanding workloads.
Caterpillar is applying lessons from automating mining operations to a very different problem: getting AI into regular businesses. The company says the hard part isn't just the technology, it's redesigning workflows around it. A useful reminder that AI deployment is starting to look more like an operations problem than a software problem.
Consulting firms are starting to feel the pressure from companies using AI to bring more work in-house, including coding and implementation. The shift could hit one of consulting's oldest advantages: charging for large teams of people to do work that software can increasingly handle.
AI traffic is becoming a much bigger part of internet activity as agents and automated systems make more requests without a human sitting behind every click. That's creating a different kind of web traffic problem: companies now have to think about machines as a major audience for their websites, APIs, and services.
Jensen Huang thinks AI has hit its inflection point: models are no longer just impressive, they're doing work companies will actually pay for. Nvidia just posted $96.2B in quarterly revenue, up 106% year over year, and expects another 70% growth next fiscal year.
Today’s AI Tools to Try:
FlintAI: Bring AI agents into Slack, Teams, Discord, and other chat apps without rebuilding your setup.
X1: Turn an idea into a real iPhone app with AI, without starting from scratch.
Krisp: Remove background noise from calls and automatically transcribe meetings.
Fluency: Automate ad setup, launches, budget pacing, and reporting from one place.
