AI News Summary — Week 22.2, 2026

A landmark week in AI. Anthropic just became the most valuable AI company in the world, OpenAI achieved a genuine mathematical breakthrough, and then... the lights went out. Here's the full story.

Anthropic Hits $965B Valuation — The New AI King

In a remarkable turn of events, Anthropic has overtaken OpenAI to become the world's most valuable AI company. The company raised $1.5 billion in Series H funding at a post-money valuation of$965 billion — surpassing OpenAI's reported $940 billion valuation.

What This Means

This isn't just a funding milestone — it's a validation of Anthropic's strategy:

• Claude is winning enterprise: The company's focus on safety, reliability, and enterprise-grade features is paying off
• The compute war is on: Anthropic's massive infrastructure deals (SpaceX, Amazon, Google, Microsoft, NVIDIA) totaling over $50 billion signal serious intent
• Talent is following the money: The steady stream of OpenAI departures to Anthropic (Karpathy, Schulman, Nordeen) now has a financial catalyst

With the new capital, expect Anthropic to accelerate its compute buildout and product expansion. The company is no longer the challenger — it's the leader.

Claude Opus 4.8: The Most Capable Claude Yet

Anthropic unveiled Claude Opus 4.8 this week — a meaningful upgrade over 4.7 with improvements across coding, agentic tasks, and professional workflows.

Key Features

• Better judgment: Opus 4.8 is notably more reliable at catching its own mistakes and flagging uncertainties. Early testers report it's "sharper" in its decision-making
• Dynamic workflows: A new Claude Code feature allowing the model to tackle massive tasks — planning work and running hundreds of parallel subagents in a single session
• Effort controls: Users can now control how much effort Claude puts into a task, similar to reasoning effort settings
• Fast mode 3x cheaper: High-speed Opus is now available at 1/3 the previous cost

Performance Highlights

• Super-Agent benchmark: Opus 4.8 is the only model to complete every case end-to-end
• CursorBench: Outperforms prior Opus models at every effort level with more efficient tool calling
• Online-Mind2Web: 84% score — the strongest computer-use and browser-agent model tested
• Honesty: ~4x less likely than Opus 4.7 to allow flaws in code to pass unremarked

Same price as before. If you're on Opus, this is an upgrade worth making.

OpenAI Solves an 80-Year-Old Math Problem

In what may be the most significant AI mathematics achievement to date, OpenAI's internal reasoning model has disproved a central conjecture in discrete geometry.

The Breakthrough

The model tackled the planar unit distance problem — an 80-year-old problem from mathematician Paul Erdős asking how many equal-sized lines can connect dots on an infinite sheet of paper. The AI-generated proof has been submitted to a top mathematics journal.

This isn't just benchmark-beating — it's genuine new mathematical knowledge. The AI didn't just find an answer; it found a proof that mathematicians are calling "the first AI proof worthy of math's top journal."

Why This Matters

For decades, AI systems have excelled at pattern recognition and prediction. Proof generation requires something different: logical reasoning, creativity in constructing arguments, and the ability to verify correctness. This result suggests we're entering a new era where AI can contribute to the frontier of human knowledge — not just replicate it.

Mathematicians are "stunned" and "amazed." That's worth paying attention to.

OpenAI Suffers Major Outage

Just as we were writing this post, OpenAI experienced a widespread service disruption affecting nearly every product:

• ChatGPT (web and mobile)
• OpenAI API
• DALL-E image generation
• Codex coding assistant
• Sora video generation
• Login system

The outage hit on May 29 — the same day as this post. At the time of writing, OpenAI had not issued an official statement on the cause or resolution timeline.

Why Developers Should Care

This isn't an isolated incident. StatusGator data shows OpenAI services in "warning" status nearly every day in May 2026. For developers building production applications on OpenAI's API, this raises legitimate questions about reliability and the need for fallback strategies.

The irony: just as OpenAI achieves a mathematical breakthrough, their systems go dark. A reminder that infrastructure matters as much as intelligence.

What's Hot This Week

• Anthropic dethroned OpenAI: $965B valuation makes Claude's maker the new AI leader
• Math breakthrough: OpenAI AI disproves 80-year-old Erdős problem — first proof worthy of top journal
• Claude 4.8 upgrades: Dynamic workflows, effort controls, and better judgment — same price
• Reliability questions: Major OpenAI outage raises concerns about API dependency

That's a wrap for Week 22. Stay tuned for more developments!

Sources: PitchBook, Anthropic, OpenAI, Scientific American, New Scientist, News9live