Uber Burned Its Annual AI Budget in Just a Few Months
An unexpected turn at Uber: the money allocated for AI for the entire 2026 budget year ran out by April. It all comes down to the adoption of Anthropic's Claude Code tool. Initially, 5,000 engineers got access, and soon almost the entire team started actively using AI to write code.
As a result, 95% of developers regularly work with neural networks, and up to 70% of the code is created with their help 🤩
The problem turned out to be the price: the company pays per request. Active developers can 'burn' up to $2,000 per month, and inside Uber they even held competitions for AI usage. As a result, expenses spiraled out of control, but no one is planning to stop 🚀
Read also
Google Disables AI Editing for Google Earth Two Days After Launch
Google introduced a feature that allowed users to use AI to literally draw events onto real satellite images. Users managed to create images of the destroyed Eiffel Tower, a huge crater near the pyramids, and dozens of other completely fictional scenes. Within 48 hours, Goo…
Former FBI agent stole nearly $1 million and asked ChatGPT for help fleeing to Europe
A major scandal is unfolding in the US. According to investigators, a former FBI employee transferred nearly $925,000 to himself from crypto wallets he accessed during an investigation. Before his arrest, he consulted ChatGPT, asking how to move to Europe with a million dollars and where best to st…
South Korea's most expensive divorce due to AI boom completed
One of the most high-profile divorce cases in recent years has concluded in South Korea. SK Group Chairman and billionaire Chey Tae-won, whose fortune skyrocketed thanks to the AI boom, was ordered by the court to pay his ex-wife approximately $645 million 💰. The reason for the long-standing dispu…
Anthropic to Pay Millions to Harry Potter Publisher
Anthropic will pay $1.5 billion in the largest copyright agreement in history. One of the main beneficiaries will be Bloomsbury, the publisher of Harry Potter books📖 Until now, AI companies claimed they could train models on copyrighted texts without permission...