OpenAI targets $1T, Claude writes 80% of its own code, and Prometheus raises $12B.

The AI industry is crossing two massive thresholds this week: the race for public market capital and the dawn of recursive self-improvement. OpenAI has confidentially filed for an IPO targeting a $1 trillion valuation, just as Anthropic warns that its AI is now writing 80% of its own codebase. Meanwhile, Jeff Bezos' Prometheus just secured $12 billion to build an "artificial general engineer," and Apple finally brought Siri into the generative era.

Anthropic has crossed a critical threshold in AI development: recursive self-improvement. In a stark blog post this week, the company revealed that as of May 2026, Claude now authors more than 80% of the code merged into Anthropic's own codebase. This means the AI model is actively building the next generation of itself. Engineers are now shipping eight times more code per quarter than they did in 2024, entirely due to Claude's assistance.

The revelation accompanied the public launch of Claude Fable 5, Anthropic's first generally available "Mythos-class" model. Anthropic had previously restricted this class of models to a select group of enterprise and government partners due to its advanced cybersecurity capabilities. Fable 5 is now available to enterprise and paid users, backed by new safety classifiers that block high-risk queries in areas like biology and cyber warfare.

The company is using this milestone to call for a global pause in frontier AI development, warning that recursive self-improvement could lead to uncontrollable capability jumps. But the warning comes at a complicated time — Anthropic just confidentially filed for an IPO, and it recently signed a massive partnership with India's Tata Consultancy Services (TCS) to deploy Claude to 50,000 associates. The tension between scaling enterprise adoption and managing existential risk has never been higher.

That is how much more code Anthropic's engineers are shipping per quarter today compared to 2024 — not because they hired more engineers, but because Claude now writes over 80% of the code that gets merged into Anthropic's own codebase. To put it plainly: the AI is building the next version of itself. For any business leader still treating AI as a productivity tool rather than a production system, this number is the clearest signal yet that the calculus has fundamentally changed.

Source: Anthropic

The private funding race is spilling onto Wall Street. OpenAI has confidentially filed its S-1 with the SEC, targeting a valuation of up to $1 trillion in an IPO that could come as early as late 2026. The move follows a recent funding round that valued the company at $852 billion. With Anthropic also filing confidentially after closing a round at a $965 billion valuation, and Elon Musk's SpaceX/xAI merger going public, the market is bracing for the three largest tech debuts in history as these companies seek the massive capital required to build out AI infrastructure.

Prometheus, the industrial AI startup co-led by Jeff Bezos and Vik Bajaj, has raised $12 billion in Series B funding at a $41 billion valuation. Backed by JPMorgan, Goldman Sachs, and BlackRock, the company is building an "artificial general engineer" designed to handle complex physical manufacturing tasks — like designing jet engines from end-to-end. For businesses in heavy industry, Prometheus signals that generative AI is finally moving beyond text and code into physical supply chains and manufacturing.

Apple finally delivered its answer to the AI assistant wars at WWDC 2026, unveiling a completely rebuilt "Siri AI" powered by the next generation of Apple Intelligence. The new Siri features onscreen awareness, allowing it to take actions based on what the user is currently looking at, and deep personal context understanding across apps like Messages, Mail, and Photos. For developers and marketers, Siri's new ability to orchestrate actions across third-party apps via the Spotlight index fundamentally changes how users will interact with iOS 27.

A new threat landscape report from CrowdStrike reveals that China-linked hacking groups are now responsible for 58% of all state-sponsored cyberattacks targeting the technology sector. The campaigns are heavily focused on intellectual property theft and supply chain access, specifically targeting U.S. AI companies to close the technology gap. For enterprise IT leaders, the report makes clear that AI models and training data are now the primary targets of nation-state cyber warfare.

NotebookLM — Google just dropped a massive upgrade to its research tool, rebuilding it on Gemini 3.5 and Antigravity. It's now a fully agentic research assistant that can search for sources, show reasoning steps, and run code.

Recall — A personal AI knowledge base that just updated to support Opus 4.8 and GPT-5.5. It automatically summarizes, categorizes, and links information from articles, videos, and podcasts into a private graph.

Customer.io — This messaging platform just rolled out deep AI automation capabilities for cross-channel marketing, making it easier for startups to build complex, personalized user journeys without manual setup.

Anthropic's detailed breakdown of how Claude now writes 80% of the company's code. This isn't just a technical flex — it's a profound look at the mechanics of recursive self-improvement. The post includes data on session success rates and explains how their engineers are shipping 8x more code than two years ago. Essential reading for any technical leader trying to understand where AI development speed is heading next.

If you only do one thing this week: read the Anthropic post on recursive self-improvement. It is the clearest signal yet that the pace of AI advancement is about to decouple entirely from human engineering constraints.