A Bankrupt Airline's 40 Years of Internal Emails Just Became a $12.5M Bidding War.

Turns out, AI is good for more than writing emails and generating LinkedIn posts. This guide is packed with 200+ actionable ways to build income with AI.

Inside you'll discover:

AI is changing how people work and how people earn, far beyond the simple “write me an email” prompt. Discover the possibilities and download the free guide to start cashing in today.

Get The Free Guide

TODAY IN AI

1  Nvidia plans to raise server prices more than 15% starting in 2027, joining Amazon and Apple in a broader wave of AI-driven price hikes.

The chipmaker reportedly plans to raise prices by more than 15% on servers shipping in 2027, including its Vera Rubin and Grace Blackwell systems, according to Bloomberg. The decision traces back to the ongoing memory chip shortage, the same root cause behind Amazon's and Apple's price hikes earlier this year. Nvidia joining the list confirms the shortage is now a cross-industry cost problem rather than something isolated to consumer electronics.

2  Anthropic's most powerful model, Claude Mythos 5, is now available to Enterprise customers through a dedicated cybersecurity product.

Claude Mythos 5 now powers Claude Security, letting Enterprise customers point the cybersecurity-savvy model at codebases, scan for vulnerabilities, and generate suggested fixes. The rollout has been deliberately slow for safety reasons. Until now, Mythos was only accessible through Anthropic's partnership program, Project Glasswing. Widening access to a model this capable, gradually and through a security-specific product first, is a notably more cautious rollout pattern than most frontier model launches this year.

3  Legal AI startup Harvey released its first post-trained model, built on top of a Chinese open-weight model rather than its own investor's technology.

Tenet is built for long-horizon legal work and optimized for token efficiency. The model is post-trained on top of Moonshot AI's Kimi K3, a decision that signals an ongoing preference for open-weight models across the industry. It is especially notable given that OpenAI has been an investor in Harvey since 2022. Building your flagship product on a competitor's open-weight model, rather than your own backer's, is the kind of decision that speaks louder than any public statement about where the actual performance and cost tradeoffs currently sit.

PRESENTED BY MORNING BREW

Morning Brew delivers the biggest stories in business, finance, and tech in about 5 minutes — with just enough personality to keep things interesting.

Join 4,000,000+ professionals who start their mornings a little smarter.

Check it out

FROM THE FRONTIER

The setup. Spirit Airlines officially went out of business in May, after 34 years. While rival airlines quickly scooped up its airport slots, tech companies have been bidding on an asset they find far more valuable: corporate records dating back to 1986.

What's actually in the trove. Spirit's data includes documents, workflows, spreadsheets, 100 million emails, 500 million Microsoft Teams messages, 7.5 billion anonymized transaction records, and much more, though notably not customer information. Mercor ($7.5M), Google ($10M), and Micro1 ($12.5M) have all submitted bids with the same motivation: using Spirit's internal records to train AI models.

Why paperwork is worth millions. Spirit's records represent a dataset that is genuinely hard to come by: an immense written record of how a large enterprise actually got real work done for years. This information, grounded in real decisions made by real human professionals rather than synthetic or scraped examples, is considerably more valuable for training AI to understand corporate workflows than generic training data found elsewhere on the open web.

Where it stands, and the lesson for everyone else. Google won the auction on Friday, but Micro1 came in with a higher bid after it closed, and it is uncertain whether the judge will accept the late offer. The whole process carries a lesson worth internalizing well beyond this one bankruptcy: be careful what you say in corporate comms. Emails and Teams messages that felt private and disposable when they were written can end up, years later, as training data for the next generation of language models, sold off in a bankruptcy auction to the highest bidder.

IN THE KNOW

🤖 Translating Claudish. The internet is so fed up with Claude's talking style that one professor actually built a Claude translator. Paste in lengthy Claude responses and it converts them into plain English, or vice versa.

🥇 Model ranking. Prominent creator Theo Browne ranked his 15 favorite AI models in order. People must agree with its accuracy: the post racked up 2M views and 13,000 likes. Check out the ranking.

🖌 Gettin' artsy. Someone hooked ChatGPT up to a receipt printer, letting it create and print art whenever it wants. Its first piece of work, Signal Garden, got 3.5M views for creatively blending technology and nature.

🐲 Soaring around. AI-generated video has not just gotten insanely realistic, it has also become incredibly easy to direct. This viral clip shows a dragon soaring around the Golden Gate Bridge, and the entire shot was created by drawing a single line across an image.

📸 No more backgrounds. ChatGPT can now create transparent images, a small but mighty update if you use GPT Image for design work. The update is only available through the API for now.

PROMPT STATION