🌟 Editor's Note: Recapping the AI landscape from 05/12/26 - 05/18/26.
🎇✅ Welcoming Thoughts
Welcome to the 43rd edition of KPAI Weekly.
What’s included: A weekly winner, AI industry impacts, Interview highlight, and Startup Spotlight.
May switch to Codex once Claude Code changes billing structure.
There was a humanoid vs human assembly line competition this week that the human narrowly won over 10 hours (12.9k > 12.7k).
Building everything you do into SOP’s continues to be the best way to automate your work.
Let’s get started—plenty to cover this week.
👑 This Week’s Winner: Anthropic // Claude
Anthropic covered more ground than anyone this week. Between a $200M global health commitment, a new small business product, and a massive enterprise rollout with PwC, Claude showed up in three completely different markets in seven days. Here's the recap:
$200M Gates Foundation Deal: Four-year partnership for global health, education, and economic mobility - vaccine screening, AI tutoring in sub-Saharan Africa, agricultural programs. Dwarfs OpenAI's $50M Gates pact from January.
Claude for Small Business: Plug-and-play connectors for QuickBooks, PayPal, HubSpot, Canva, DocuSign, and more. Fifteen workflows out of the box.
PwC Global Rollout: PwC deploying Claude to hundreds of thousands of professionals. 30,000 employees trained, new CFO unit built around it.
Anthropic also Acquired Stainless, the company behind its developer toolkits, and announced metered credit pools for heavy programmatic usage starting June 15. Anthropic is going to start charging for agentic/programatic use in a way that closer resembles API costs than general chat subscription. No good.
From Top to Bottom: Open AI, Google Gemini, xAI, Meta AI, Anthropic, NVIDIA.
⬇️ Field Update
Who’s moving, who’s stalling, and who’s climbing: Ordered by production this week.
🟣 Google // Gemini
Blackstone AI Cloud JV: Google and Blackstone are launching a joint AI cloud company with $5B in Blackstone equity. It rents out Google's custom AI chips as a service, with 500 megawatts of power online by 2027. Wow! Last minute deal to close the week. Google getting more involved in compute is a positive.
🟢 OpenAI // ChatGPT
Personal Finance in ChatGPT: ChatGPT Pro users can now connect bank accounts, credit cards, and brokerages from 12,000+ institutions and track spending, net worth, and investments in one place. OpenAI says 200 million users already ask financial questions monthly. Smart consumer play. Guessing this would’ve released sooner if not for security validation.
🔴 xAI // Grok
Grok Build Coding Agent: xAI shipped Grok Build, a terminal-based AI coding assistant for SuperGrok Heavy subscribers ($300/mo). It plans, writes, and reviews code with parallel automation. Positioned against Claude Code and Codex. Musk has admitted xAI "fell behind" on coding. I think xAI can build a powerful coding tool, I don’t think it will catch Claude Code or Codex anytime soon.
⚪️ NVIDIA
Jensen on Air Force One: Trump personally called Jensen Huang to join his Beijing trade delegation. Huang flew to Alaska to board Air Force One alongside Tim Cook, Elon Musk, and Larry Fink for the Trump-Xi summit.
🔵 Meta // Meta AI
8,000 Layoffs Begin This Week: Meta cutting 10% of its workforce starting May 20, plus scrapping 6,000 open roles. Funding a $125-145B AI buildout. WIRED: some employees are hoping to get laid off for 16 weeks of severance. This has been expected for a while, but I can’t imagine constant layoff cycles / distractions are conducive to positive growth.
🚑 Impact Industries 📢
Healthcare // Isomorphic Labs $2.1B Series B
Isomorphic Labs, the Alphabet spinout led by DeepMind CEO Demis Hassabis, closed a $2.1B Series B led by Thrive Capital — the second-largest biotech round ever behind Altos Labs' $3B debut in 2022. New backers include MGX, Temasek, CapitalG, and the UK Sovereign AI Fund. The capital will scale its AI Drug Design Engine and push the first AI-designed drugs into human trials by year-end 2026. PitchBook's Ben Zercher noted Isomorphic "appears to be operating entirely outside of this paradigm" — the rare AI-bio company allowed to stay opaque on disease targets at this round size due to its DeepMind halo. If those trials materialize on schedule, it becomes the highest-stakes test yet of whether AI can genuinely compress drug discovery timelines.
Read the Story
Marketing & Advertising // TikTok Ads MCP Server
At TikTok World on May 13, the company launched an Ads MCP Server and developer toolkit that lets brands connect their own AI agents directly into TikTok's ad stack. Campaign creation, bidding, and optimization can now run end-to-end without a human touching the dashboard. It follows similar MCP server launches from Meta and Amazon, but TikTok's is the broadest agentic ad surface yet exposed by a major consumer platform. Global head of product marketing Jose Villalobos framed it as the next step in making TikTok's ad business fully programmable. With all three major social ad platforms now offering MCP-based agent access, the infrastructure for fully automated campaign management is in place across the industry.
Read the Story
💻 Interview Highlight: Krishna Rao (Invest Like the Best Pod)
Interview Outline: Krishna Rao discusses the financial and operational realities of scaling a frontier AI lab. He breaks down Anthropic's multi-cloud compute strategy , the "Cone of Uncertainty" that governs exponential growth planning , and how the returns to frontier intelligence unlocked a revenue run rate leap from $9 billion to over $30 billion in a single quarter.
About the Interviewee: Krishna Rao is the CFO of Anthropic, joining the company during its Series D round. He previously worked in the private equity group at Blackstone and helped lead Airbnb's landmark restructuring and financing during the pandemic.
Interesting Quote: "If you buy too much compute, you go out of business. If you buy too little compute, you can't serve your customers."
Condensed Interview Highlight — Krishna Rao (Invest Like the Best Podcast)
1. Patrick: Can you explain your framework of the "cone of uncertainty" in your growth planning?
Krishna Rao: In exponential growth, tiny growth shifts compound into wildly different outcomes over a year or two. Humans naturally think linearly and incrementally, which is a paradigm I had to completely break. Traditional quarterly forecasting fails here; we plan around dynamic scenarios and instantly update our priors when model shifts break our assumptions.
2. Patrick: Why are the economic returns to the bleeding-edge frontier remaining so high for enterprises?
Krishna Rao: Each model generation unlocks an entirely new set of multi-dimensional capabilities. It opens new TAM and use cases, like long-horizon agentic tasks and tool use. To show the velocity: we went from a $9 billion revenue run rate to north of $30 billion in a single quarter because enterprise clients instantly scale up token consumption when capabilities jump.
3. Patrick: How is the finance team deploying Claude internally to run corporate operations?
Krishna Rao: If we aren't super-users pushing limits, we can't expect customers to do it. All of our legal entities' statutory financial statements are now produced by Claude. We have a library of over 70 finance skills, including a Monthly Financial Review skill that generates reports 95% ready. What used to take hours now takes 30 minutes, freeing us to focus entirely on strategy.
4. Patrick: When you dropped Mythos, it caused waves due to its offensive capabilities. What was that decision process like?
Krishna Rao: Mythos is highly capable across many dimensions, but its cybersecurity capabilities spiked drastically. A prior model found 22 vulnerabilities in an open codebase; Mythos found 250. Because of that, we launched it in a phased way to a controlled group so it could be deployed to patch infrastructure defensively rather than offensively.
5. Patrick: You've raised an immense envelope of capital. How do sophisticated investors perceive your business model?
Krishna Rao: We’ve raised $75 billion since I joined, plus another $50 billion in commitments via Google and Amazon infrastructure deals. Investors initially tried to fit us into a standard SaaS mold. But our net dollar retention rate is over 500% annualized, and we serve nine of the Fortune 10. We don't raise money to fund current losses; we raise it to insulate the pipelines required to stay at the frontier.
📚 Startup Spotlight
Mark II
Mark — The smart bookmark that brings AI to physical books.
The Problem: Reading real books feels great, but keeping track of insights sucks. When you hit a killer quote, you either have to fold the page, ruin the book with a yellow highlighter, or stop reading entirely to manually type it into your phone. It kills your reading flow and leaves your best knowledge trapped on paper.
The Solution: A slim, sleek physical bookmark packed with an onboard optical scanner and microphone. You just slide Mark over a line of text to instantly digitize the quote, or tap it to record a quick voice note. Everything automatically syncs to a companion app that indexes, categorizes, and lets you search through your physical reading notes using AI.
The Backstory: Built specifically for analog purists who refuse to switch to a Kindle, Mark II bridges the gap between tactile paper and digital organization. The startup recently locked in a new round of equity financing to ramp up hardware manufacturing, turning physical bookshelves into fully searchable, AI-driven second brains.
My Thoughts: I’m not much of a book reader, but I’ve heard great things about this for those that are. Seems like a cool product.
“It’s not likely you’ll lose a job to AI. You’re going to lose the job to somebody who uses AI”
- Jensen Huang | NVIDIA CEO
Till Next Time,
Noah from KPAI