Opus 4.8, Anthropic Hits $900B, Meta Surveils Staff: AI News (May 29, 2026)

1. Claude Opus 4.8 Released

Anthropic dropped Opus 4.8 on May 28. Agentic coding jumped from 64.3% to 69.2% (beating GPT-5.5's 67.1%). Computer use hit 83.4%, making it the best browser agent model available. Knowledge work rose from 1753 to 1890. The model is 4x less likely to miss flaws in its own code compared to 4.7.

New features: effort control (choose Low, High, or Max thinking per task), dynamic workflows in Claude Code for larger autonomous projects, and fast mode now 3x cheaper. Pricing unchanged at $20/mo (≈₹1,860/mo) Pro and $100/mo (≈₹9,300/mo) Max. Mythos 1, described as "even larger and more powerful," is teased for October.

I've updated our Claude review and Composer 2 vs Claude Sonnet comparison with Opus 4.8 details.

Full coverage: Opus 4.8 deep dive

2. Anthropic Closes $30B Round at $900B+

Anthropic's $30 billion round officially closed this week, co-led by Sequoia, Dragoneer, Altimeter, and Greenoaks (roughly $2B each). Peter Thiel's Founders Fund, General Catalyst, Microsoft, and NVIDIA also participated. At $900B+, Anthropic is now the world's most valuable private AI company, surpassing OpenAI's $852B March valuation.

The speed: $380B in February, $852B (OpenAI's mark) in March, $900B+ in May. Four months to flip the valuation lead.

Full coverage: Anthropic $30B round + SpaceX IPO

3. KPMG Deploys Claude to 276,000 Employees

KPMG launched "Digital Gateway Powered by Claude" across 138 countries, integrating Claude Cowork and Managed Agents into their core client delivery platform on Microsoft Azure. KPMG joins Deloitte (470K employees), PwC (276K), and BlackRock ($11.5T AUM) as major enterprise Claude deployments. Combined reach: over 1 million enterprise users across the Big Four consulting firms.

Full coverage: KPMG + DeployCo

4. OpenAI Launches $4B DeployCo

OpenAI's counter-move: DeployCo, a majority-owned consulting subsidiary backed by $4B from 19 investment firms. It helps enterprises deploy ChatGPT and GPT-5.5. The tension: OpenAI is now competing with the same consulting firms (KPMG, Deloitte, Accenture) that could be selling ChatGPT Enterprise. Anthropic partners with consultants. OpenAI is becoming one.

Full coverage: KPMG + DeployCo

5. Meta Tracked Employees to Train AI, Then Fired 8,000

Leaked audio from an April all-hands captures Mark Zuckerberg explaining that Meta installed keystroke, mouse, and screenshot tracking across Gmail, GChat, VSCode, and Metamate on employee computers to train AI models. The audio leaked May 19. On May 20, 8,000 employees received termination emails starting at 4 AM. Employees posted fliers and signed petitions. Meta's CISO responded by threatening to fire leakers.

Full coverage: Meta surveillance + OpenAI math proof

6. Pope Leo XIV Releases First AI Encyclical with Anthropic

Pope Leo XIV presented "Magnifica Humanitas," signed 135 years to the day after Rerum Novarum (workers' rights during the Industrial Revolution). He co-presented alongside Anthropic co-founder Christopher Olah. The Vatican chose Anthropic over Google and OpenAI, a direct endorsement of safety-first AI development.

Full coverage: Pope encyclical + $900B valuation

7. Gates Foundation + Anthropic Commit $200M

A $200M, four-year partnership for AI in healthcare (vaccine screening, malaria/TB forecasting), education (K-12 tutoring in India and sub-Saharan Africa), and agriculture (smallholder farming AI). Four times larger than Gates' $50M OpenAI partnership. African language training data being released publicly.

Full coverage: Gates Foundation partnership

8. Karpathy Joins Anthropic + Gemini 3.5 Flash GA

Andrej Karpathy, OpenAI co-founder #11 and former Tesla AI Director, joined Anthropic's pre-training team. His role: build a team that uses Claude to accelerate training the next Claude. Separately, Gemini 3.5 Flash API went GA at $1.50/$9 per M tokens (≈₹140/₹837) with a 1M context window, the best price-to-context ratio in the API market.

Full coverage: Karpathy + Flash GA

9. SpaceX IPO: June 12 on Nasdaq

SpaceX confirmed its IPO timeline: roadshow June 4, pricing June 11, trading June 12 under SPCX on Nasdaq. Target: $75B raise at $1.75T valuation. 30% routed to retail investors via Robinhood, Fidelity, and Schwab. If it hits target, it surpasses Saudi Aramco 2019 as the largest IPO in history. The AI connection: Anthropic pays SpaceX $1.25B/mo for Colossus 2 compute through 2029.

Quick Hits

• OpenAI model disproves 80-year geometry conjecture. First AI to solve a prominent open math problem autonomously. Fields medalist Tim Gowers called it "a milestone."
• AI beats average humans on creativity tests. 100K-person study from Université de Montréal. GPT-4 and Gemini outperform average human. Top 10% of humans still far ahead.
• China restricts AI researcher travel. Bloomberg reports widening overseas travel limitations for researchers at Alibaba and DeepSeek.
• Google May 2026 core update continues. Rolling out through roughly June 4. Don't judge ranking changes until it completes.
• OpenAI named Gartner Magic Quadrant leader for Enterprise AI Coding Agents. Cursor and Claude Code notably absent.

The Anthropic Pattern

This was the most one-sided week in the AI industry's history. Anthropic dominated every category: product (Opus 4.8), valuation ($900B+), enterprise (KPMG, 1M+ users), partnerships (Gates $200M), talent (Karpathy), and institutional trust (Vatican). Six major wins in seven days.

The contrast with Meta is instructive. While Anthropic hired researchers, released public data, and co-authored papal documents, Meta surveilled its own engineers and fired 8,000 of them. Two companies building AI. Two very different approaches. The market is telling you which one it prefers.

What I'm Watching Next Week

• Microsoft Build (June 2-4). Any Claude or Gemini integration announcements would signal enterprise fragmentation away from OpenAI exclusivity.
• Apple WWDC (June 8). Siri overhaul and on-device AI updates could reshape the consumer AI landscape.
• SpaceX IPO roadshow (starts June 4). Investor reception sets the tone for OpenAI (September) and Anthropic (October) IPOs.
• Google core update completion (roughly June 4). First reliable post-update ranking assessment for all publishers.
• Opus 4.8 full benchmark. I'm running the complete test suite this week to update Claude review scores.

This roundup covers May 23-29, 2026. Previous roundup: May 22. All prices at ≈₹93/USD.

More from RawPickAI: Claude Review · Composer 2.5 Review · Transparency Index · 2026 AI Tools Reality Check · Compare AI Tools