Nadella: Own AI ‘Token Capital’ or Lose Your Industry
ALSO: 40K AI layoffs in May, AI cracks 18 rare diseases
Welcome to Guru99 AI Report!
Top Story: Hey there — this week, Nadella says the smartest AI model won’t save you; owning your loop will. Plus: record profits meet mass layoffs, AI cracks 18 rare diseases, and Nvidia bets big on world models. Let’s dive in.
🧠 Nadella’s Warning: Build ‘Token Capital’ or Cede Everything
Brief Buzz:
Microsoft CEO Satya Nadella dropped a lengthy weekend essay arguing the AI race isn’t about owning the smartest model — it’s about owning the loop. His pitch: pair human capital with proprietary token capital, or risk ceding your industry’s value to a handful of giant models.
- Every firm now has two assets: human capital — your people’s judgment, relationships, and pattern-recognition — and token capital, the AI systems you build and own. One amplifies the other; neither replaces it.
- The edge isn’t picking the smartest model, but building a learning loop that compounds your data and know-how, so you can swap base models without losing hard-won expertise.
- Nadella’s blunt warning: if a few foundation models swallow all the value, “the political economy will simply not tolerate it.”
💡 Why Should You Care?
For any business betting on AI, the lesson is ownership: rent the model, own your data loop. It’s also a rare big-tech admission that AI concentration could spark backlash.
💥 Record Profits, Mass Layoffs: AI’s Powder Keg
Brief Buzz:
Tech is posting record profits while cutting jobs at a blistering pace — and blaming AI. But a growing chorus of skeptics, from top VCs to CEOs, argues the tech is a convenient scapegoat for pandemic-era overhiring, not the real cause.
- Tech layoffs hit their highest monthly total in two years in May — nearly 40,000 cuts — with AI the most-cited reason across industries, per Challenger, Gray & Christmas.
- Tracker TrueUp counts ~150,000 tech job losses in 2026 so far — about 974 a day, a pace 44% faster than last year.
- Skeptics aren’t convinced: VC Marc Andreessen calls AI the “silver bullet excuse” for cuts that are really about bloated, overhired companies.
- Meanwhile, AI insiders mint fortunes: chipmaker Cerebras debuted up 68% on the Nasdaq, and Meta cut ~8,000 jobs just after Zuckerberg’s record $170M Miami mansion.
- Firms like Block, Cloudflare, and Atlassian saw stocks rise after pinning cuts on AI — even as Cloudflare’s revenue hit record highs.
💡 Why Should You Care?
Whether or not AI is doing the work, it’s become the default excuse for cuts. Workers should question the “AI did it” line — and watch a wealth gap that history says rarely stays quiet.
🔍 Facebook’s AI Mode Mines Public Posts for Answers
Brief Buzz:
Meta just rewired Facebook search. Its new AI Mode skips the list of links and hands you a single, conversational answer — stitched together from public posts, Groups, and Reels. It’s Google’s AI Overviews logic, except the “experts” are whoever happened to post publicly.
- AI Mode lives inside Facebook search, powered by Meta’s Muse Spark model — ask in plain language, get a synthesized answer instead of a link list.
- It pulls from public posts, Groups, and Reels, mirroring Meta’s Reddit-style Forum app and its “Ask” tab.
- The catch: answers come from everyday users, not vetted sources, raising the same misinformation risk already flagged for Google’s AI results on Reddit.
- Meta also shipped AI photo and video editing — collage cutouts, transitions, and presets that swap clothes, hairstyles, even sports-team jerseys via “Wear It.”
- It’s part of a monetization push: paid subscriptions from $3.99/month, with more AI tiers reportedly coming.
💡 Why Should You Care?
If you build or rely on AI search, this is generative answers drawn from unvetted social chatter at scale. For everyone else: convenient, but treat Facebook’s AI answers like a stranger’s tip in the comments.
🧬 AI Helps Crack Rare Diseases — and Manage Them
Brief Buzz:
Two healthcare AI milestones landed this week. OpenAI’s reasoning model helped doctors at Boston Children’s Hospital surface 18 new diagnoses in cases stuck for years, while Google’s AMIE matched primary-care physicians at managing ongoing disease — signaling AI’s shift from lab demo to clinical teammate.
- OpenAI’s o3 Deep Research model reanalyzed 376 unsolved cases, surfacing leads for 18 diagnoses — a 4.8% gain on cases experts had already exhausted.
- Crucially, the AI never diagnosed anyone: it generated evidence-linked hypotheses that physicians verified and confirmed in certified labs.
- One patient, Kyra, finally got an answer after a nearly 20-year diagnostic odyssey.
- Separately, Google’s AMIE outperformed 21 doctors on treatment-plan precision and guideline alignment across 100 simulated multi-visit cases, per its Nature study.
- Both teams stress the systems are research tools — not ready for real-world clinical use yet.
💡 Why Should You Care?
Half of rare-disease patients never get a diagnosis — so tools helping experts re-mine old data could end years of uncertainty. And smarter management AI points toward more guideline-backed care, without replacing your doctor.
📊 Perplexity Finance turns AI into your free stock analyst
Brief Buzz:
Want stock research without juggling a dozen tabs or paying for Bloomberg-grade tools? Perplexity Finance, free inside the Comet browser, bundles screening, financials, earnings, and analyst targets into one AI-powered hub that explains the why behind any price move.
- Use the natural-language Screener: type plain-English filters like “low-P/E energy names,” but keep the first pass broad, since tight screens often return nothing.
- Open any ticker for Overview (news plus bull/bear context), Financials, Earnings beats and call highlights, and Analysis with Wall Street price targets.
- Ask follow-ups through the Comet assistant (top-right, not the bottom chat box, which hides the data), then click and date-check every cited source.
- Connect a brokerage through Plaid to analyze your real portfolio, and set price alerts and watchlists to track names over time.
💡 Why Should You Care?
For everyday investors, this collapses an hour of tab-hopping into minutes — but it’s a research starting point, not advice. Always verify the cited figures and dates yourself. Read more: Explore the full Perplexity Finance toolkit.
🤖 Nvidia bets big on world models with Cosmos 3
Brief Buzz:
As the AI race expands beyond chatbots, Nvidia is going all-in on the tech behind physical AI. At Computex in Taipei, it unveiled Cosmos 3, a “fully open” model that reasons and generates across text, video, images, sound, and action.
- Cosmos 3 uses a new mixture-of-transformers architecture — one transformer to reason about a scene, another to generate it — so it grasps motion and object interactions before producing video or action paths.
- It’s an omnimodel: image, video, sound, and action are treated as equal partners alongside text, rather than bolted onto a language model.
- Nvidia trained it on a huge dataset — 20 trillion tokens, nearly a billion images, and 400 million real and synthetic videos.
- The models are open-source, with two sizes available now (Super and Nano) and a real-time Edge version coming soon, so companies can run and customize them locally.
- Nvidia calls better generalization — robots adapting to unfamiliar situations — the “holy grail” of robotics, and says Cosmos 3 is a foundation toward it, not the fix itself.
💡 Why Should You Care?
Today’s robots still stumble in unpredictable settings. By open-sourcing a model that learns how the physical world behaves, Nvidia could speed up smarter home robots, warehouse machines, and self-driving cars — and let companies build on it without vendor lock-in.
📧 AI Jobs
- Micro: AI/ML Engineer — Python, LLMs →
- Respondent: Paid AI research participant studies →
- JobCopilot: AI-powered job application automation →
Hey! I’m Krishna Rungta
Founder of Guru99.com, Editor-in-chief & Technology Expert
Was this email forwarded to you? Sign up for free here.
