Remember how fast this is.
The AI startup triples its valuation from March as run-rate revenue tops $5 billion, driven by enterprise demand and its Claude Code developer tool.
The ChatGPT-maker buys a product-experimentation startup and reshuffles its executive ranks as it builds out a dedicated applications unit under new CEO Fidji Simo.
A rare public step-back: as clips of the drive-through AI misfiring rack up millions of views, Taco Bell's technology chief concedes mixed results and says human staff will keep taking orders at the busiest locations.
Claude Free, Pro, and Max users must decide by October 8 whether to allow their conversations to be used for model training, with a five-year data retention period for those who do not opt out.
Microsoft AI, led by Mustafa Suleyman, ships MAI-Voice-1 and MAI-1-preview — the first major models built entirely in-house — signaling a strategic shift away from sole dependence on OpenAI.
The company says it disrupted a data extortion campaign that used its AI to target at least 17 organizations, with ransom demands exceeding $500,000.
The AI chip maker's data center revenue rose 56 percent year over year to $41.1 billion, and shares slip after hours as data-center revenue comes up short of the most optimistic forecasts.
The Weekly Replay · free by email
One email each Sunday: the week's replayed AI news, with the one-year-later annotations included. Written like it's breaking — dated like it isn't.
Free · double opt-in · unsubscribe anytime · privacy
Google's new image model brings character consistency across edits, multi-image blending, and precise local changes from plain-language prompts — each output carrying an invisible SynthID watermark.
The controlled pilot gives 1,000 Max plan subscribers an AI agent that can fill forms, manage calendars, and navigate websites, while Anthropic races to contain prompt-injection attacks that succeeded 23.6% of the time in testing.
The $3,499 developer kit and $2,999 production module for orders of more than 1,000 chips bring 7.5x the AI performance of the prior generation, enabling humanoid robots and edge devices to run generative models locally.
Elon Musk releases Grok 2.5 model weights on Hugging Face under a custom license.
The deal gives Meta access to the startup's image and video generation tech as it looks to improve visual quality in its products.
In a podcast interview, Armstrong described ordering engineers to onboard with GitHub Copilot and Cursor by a weekend deadline and terminating those who didn't comply without a valid excuse.
The Chinese AI lab releases a unified model that toggles between fast answers and chain-of-thought reasoning, and says its new FP8 format anticipates the next generation of homegrown silicon.
At its Made by Google 2025 event, the company positions the phone as an AI assistant first and a handset second, with new features like Magic Cue, Camera Coach, and conversational photo editing.
The shakeup places most units under new chief AI officer Alexandr Wang, dissolves the AGI Foundations team, and halts the aggressive talent raids that marked the summer.
A data-backed study of 300+ deployments finds the failure is organizational, not technical — generic tools like ChatGPT stall because they cannot learn or integrate into workflows.
Motivated by model welfare research, Anthropic allows Claude to walk away from extreme harmful interactions as a last resort.
An internal 200-page policy manual also permitted bots to generate false medical information and demeaning statements about protected characteristics, according to Reuters.
Tomorrow’s news publishes tomorrow. It has already happened. No peeking.