Sponsored by
Good morning, {{first_name | AI enthusiast}}.
SpaceX just closed the biggest AI acquisition of the summer. Elon Musk’s rocket company officially owns Cursor, the AI coding assistant it agreed to buy for $60 billion back in April — and Cursor now plugs straight into SpaceX’s Colossus supercomputer, one of the largest GPU fleets on the planet.
It’s the clearest sign yet that the biggest AI players are consolidating around whoever controls the most compute, not just the best models. If coding tools become bargaining chips in a compute arms race, what happens to the developers caught in the middle?
Today in AI Brief:
SpaceX officially closes its $60B Cursor buyout
OpenAI’s Ultrafast mode runs GPT-5.6 Sol 14x faster
DeepSeek open-sources Harness, a free Claude Code rival
How Jennifer Aniston’s LolaVie brand grew sales 40% with CTV ads
The DTC beauty category is crowded. To break through, Jennifer Aniston’s brand LolaVie, worked with Roku Ads Manager to easily set up, test, and optimize CTV ad creatives. The campaign helped drive a big lift in sales and customer growth, helping LolaVie break through in the crowded beauty category.
SpaceX Officially Closes Its $60 Billion Cursor Acquisition
In Brief: SpaceX completed its acquisition of AI coding startup Cursor for $60 billion in stock, four months after the companies first struck the deal in April.
The Details:
Cursor now taps into the largest fleet of GPUs in the world, including SpaceX’s Colossus supercomputer, according to the company’s own description of the deal.
The purchase follows SpaceX’s earlier acquisition of xAI, making Cursor the second major AI company Musk has folded into the rocket maker in recent months.
In a staff meeting, Musk said SpaceX’s AI revenue would exceed all of its other revenue — including satellites and launches — by September.
Take Away:
Compute, not model quality, is becoming the currency AI labs fight over — and Musk just used a rocket company’s balance sheet to buy his way into coding tools. Expect rivals to start treating GPU access as a bigger moat than any single model release.
OpenAI’s Ultrafast Mode Runs GPT-5.6 Sol 14x Faster
In Brief: OpenAI began previewing Ultrafast mode, a new API tier that runs GPT-5.6 Sol up to 14 times faster than standard processing, powered by a partnership with chipmaker Cerebras.
The Details:
Ultrafast pushes throughput as high as 750 tokens per second, powered by Cerebras’ wafer-scale chips, letting the model finish a 2,500-question exam in 11 hours instead of the 78 hours its predecessor needed.
The tier runs the same model weights as standard GPT-5.6 Sol, so speed no longer means trading away intelligence — it’s purely an infrastructure choice.
Access is currently limited to select customers testing Ultrafast in coding, commerce, and support workflows, with no public pricing or launch date announced yet.
Take Away:
Model races used to be about who’s smartest; this one is about who responds first. If speed and intelligence really are decoupled, expect AI to start showing up in the most time-sensitive parts of business — trading desks, live support, security triage — where a slow answer is a useless one.
SPONSORED BY LIQUID
One Account. Every Market. No Closing Bell.
Markets don't wait for Monday. News breaks on a Saturday morning, and most traders can do nothing but watch.
Not on Liquid. Trade domestic and international equities, commodities, forex, crypto, and prediction markets — all from one account, 24 hours a day, 365 days a year. Liquid gives you access to any market, from anywhere, anytime. To us, access is arbitrage.
Getting started takes under 10 minutes: log in with Google, deposit with Apple Pay or a bank transfer, and trade from your phone or desktop — wherever you are in the world.
While everyone else is refreshing headlines and waiting for the open, you're already positioned. That's the difference between reacting to markets and actually trading them.
DeepSeek Open-Sources Harness, a Free Rival to Claude Code
In Brief: DeepSeek released Harness v0.1 in developer preview, an open-source agent harness — the runtime layer that turns a language model into an agent capable of reading files, running commands, and completing multi-step tasks.
The Details:
The project is MIT licensed and built on Cordis, a framework where every component — models, tools, sessions, sandboxes — is a swappable plugin.
Harness supports DeepSeek, Anthropic, OpenAI, Bedrock, Vertex, and Azure as interchangeable model providers, so developers aren’t locked into DeepSeek’s own models to use it.
The repo topped 126,000 GitHub stars within its first three days — a sign of just how much pent-up demand exists for a free alternative to Anthropic’s Claude Code.
Take Away:
A free, open-source agent runtime that works with any model provider undercuts the business model of paid coding agents overnight. Watch whether Harness’s rocky v0.1 growing pains slow that momentum, or whether “everything is a plugin” becomes the default way agent frameworks get built.
Everything else in AI
Google launched Gemini 3.7 Flash just three weeks after 3.6 Flash, cutting coding and web-dev prices in half while beating its predecessor on nine benchmarks.
Anthropic published its August Risk Report, raising its misalignment risk rating to “low” and disclosing an unreleased internal model, Model 2, that already beats Claude Mythos 5 on Anthropic’s own coding benchmark.
Amazon shipped Quick for Microsoft 365, an agentic AI layer that edits Word docs, builds PowerPoint slides, and drafts Outlook replies directly inside Microsoft’s own apps.