There’s not a day that goes by without me thinking, what an extraordinary time to be alive. When used wisely, AI is a remarkable force for good. Take the Mayo Clinic’s breakthrough—diagnosing nine dementia types from a single scan—proof of how AI can transform medicine and improve countless lives. Investors and hedge funds are pouring billions into AI startups, a wave that will inevitably carry both successes and failures. Yet even with the risks, I see far more promise than peril. That’s why I proudly call myself an AI Maximalist.
Sam
Hedge funds are rapidly deploying capital into AI-centric investment strategies, launching new funds and delivering strong early returns amid a broader fervor around AI.
🔍
Many of the new AI hedge funds mix long positions in AI assets with shorts in at-risk legacy industries.
One startup founder, a former OpenAI researcher, raised over $1.5 billion and saw ~47 % gain in H1.
Investors are committing to long lock-up periods, underscoring confidence in the AI investment thesis.
Both public equity and private startup sectors are key targets for fund allocations.
Mayo Clinic’s AI-based StateViewer diagnoses nine dementia types—including Alzheimer’s—from a single FDG-PET scan with high accuracy and speed.
🔍
Achieved 88% correct dementia-type identification using over 3,600 patient scans.
Delivers results nearly twice as fast and up to three times more accurate than standard methods.
Visual, color-coded brain maps clarify diagnosis, aiding clinicians without neurology training.
May improve early, precise treatment and extend advanced diagnostics to under-resourced clinics.
Demis Hassabis outlines how Gemini 2.5’s Deep Think, Genie 3’s world modeling, and Kaggle’s Game Arena benchmark propel AI toward AGI.
🔍
Gemini 2.5’s “Deep Think” mode enhances reasoning by allowing the model to plan before responding.
Genie 3 introduces immersive world models enabling AI agents to simulate real-world environments for training.
Game Arena on Kaggle offers a dynamic, competitive benchmark to test AI adaptability and strategizing.
Focus is on systems that generalize across domains, learn efficiently, and understand cause-and-effect
udge rules training AI on legally purchased books is fair use, but Anthropic still faces a December trial over use of pirated materials.
🔍
Decision rooted in transformative use, likened to “reader aspiring to be a writer.”
Over 7 million pirated books stored in central library deemed infringement.
December trial will determine damages, potentially substantial per-work fines under U.S law.
Ruling sets precedent for AI training; may influence lawsuits against OpenAI, Meta.
GitHub CEO Thomas Dohmke resigns mid-2025; GitHub now reports to CoreAI, ending its independent leadership without a successor.
🔍
Dohmke leaves to return to his startup roots, stays through end of 2025 for transition.
GitHub leadership will report directly to Microsoft’s CoreAI, led by Jay Parikh.
No new CEO will be appointed—GitHub loses standalone status.
Reflects Microsoft’s strategy to integrate AI tightly into developer tools platforms.
President Trump permits Nvidia and AMD to sell H20 (and MI308) chips to China, with the U.S. taking a 15% share of sales revenue.
🔍
Trump initially demanded 20% but accepted 15% after negotiation with Nvidia’s CEO.
The arrangement reverses an earlier export ban tied to national security concerns.
Lawmakers criticized the model as a “creative tax,” raising constitutional and security alarms.
Trump downplayed the chip’s quality, calling the H20 “obsolete,” even while approving its export.
Reddit restricts the Internet Archive to only crawling its homepage, halting access to full discussions, comments, and profiles amid concerns over AI scraping.
🔍
Reddit cites AI firms scraping sensitive data through archived pages.
Wayback Machine still allowed to archive only Reddit’s homepage.
Goal: better protect user privacy and removed or private content.
Encourages data access via paid deals and formal API agreements.
Upcoming Event:
Your opinion matters. 🤝
Forward to a friend 📩
If you find this newsletter valuable, please consider forwarding it to a friend or colleague who may also benefit from staying up to date with the latest in AI.
If you have any suggestions, questions, or ideas, feel free to reply to this email. We’d love to hear from you and are always looking to improve our newsletter.