Good morning. It's Friday, July 17, and we're covering the push for a universal AI identity standard, India's latest AI unicorn fueling the startup race, and Linus Torvalds' latest take on AI in Linux.
Know someone who'd love this? Forward it their way. First time reading? Sign up here to keep up with the future of tech.
Rather than limiting AI use, Linus Torvalds says critics should fork Linux or leave the project entirely. He argues Linux is not an anti-AI project and maintains developers are free to use AI tools despite community opposition. Torvalds says AI already proves useful for finding bugs, though it can generate bloated code, marking a notable shift from his more skeptical 2024 stance.
Apple's AI rollout in China lifts Alibaba and Baidu shares after regulators approve Apple Intelligence for the market. Alibaba confirms its Qwen model will power Apple Intelligence across iOS, iPadOS, macOS, and visionOS in China, while Baidu supports additional iPhone AI features. Hong Kong-listed Alibaba jumps 5% and Baidu gains 4% as the approval adds momentum to China's intensifying AI competition with the U.S.
TSMC pairs record earnings with a massive new U.S. manufacturing pledge as AI chip demand continues to accelerate. The chipmaker reports $40.2 billion in quarterly revenue and a record NT$706.6 billion profit, while raising its 2026 revenue growth forecast to above 40%. It also commits an additional $100 billion to Arizona fabs, bringing total planned U.S. investment to $265 billion.
Moonshot AI is preparing an open-weight model that reportedly rivals Anthropic's Opus 4.8, narrowing the gap with leading closed-source systems. The upcoming Kimi K3 is expected to feature 2–3 trillion parameters and launch within days, according to the Financial Times. Moonshot is also reportedly seeking new funding at a $31.5 billion valuation after raising $2 billion at a $20 billion valuation in May.
FRIDAY FACT
The Hard Drive That Fits in a Raindrop
One gram of a certain material can hold more data than every server farm on Earth combined, and it's been sitting inside you your whole life. More at the bottom.
VIDEO
Anthropic's Compute Problem Persists
OpenAI is winning developers with generous quotas and lower costs, while Anthropic's compute limits risk undermining its lead with Fable.
ZEITGEIST POLL
Is a "Three Mile Island Moment" Coming for AI?
In 1979, a meltdown that killed no one still stalled US nuclear power for a generation.
Is AI heading towards a similar fate? |
STANDARDS
Vint Cerf Backs Open Standard for AI Agent Identity
Internet pioneer Vint Cerf has joined Innovation Labs as an adviser to help develop an open standard for identifying AI agents across the internet, the organization announced on July 15, 2026. The group’s proposed DNSid system would tie AI agents to existing domain names using cryptographic proofs, making it easier to verify authority, establish accountability, and build trust as autonomous agents interact across different platforms.
Innovation Labs says it is already testing the standard with several unnamed hyperscalers and identity companies, positioning it as an open alternative to proprietary approaches. Cerf said widespread adoption will depend on interoperability rather than any single company’s technology, drawing a parallel to how TCP/IP became the foundation of the modern internet. → Read the full article here.
STARTUPS
Emergent Becomes India’s Second AI Unicorn in a Month
Bengaluru-based vibe coding startup Emergent has raised $300 million in a Series C funding round, reaching a $1.5 billion valuation and becoming India’s second AI unicorn in a month. The company says more than 12 million apps have been built on its platform, with 70% of users having no prior coding experience, highlighting growing demand for AI-powered software creation.
The milestone follows Sarvam’s recent $1.5 billion valuation and reflects rising investor confidence in India’s AI ecosystem, even as the country trails global leaders in chips, data centers, and frontier AI models. Industry analysts see India’s strength in building AI applications and leveraging its deep software talent, though they caution it could take another three to four years for the ecosystem to reach self-sustaining scale. → Read the full article here.
HEALTHCARE
Mayo Clinic Expands AI Across Hospital With 150 Clinical Models
Mayo Clinic has deployed more than 150 AI models across its hospital system, using the technology to streamline clinical workflows, support earlier disease detection, and reduce administrative burdens. One tool, Record Time, summarizes patient records and organizes medical histories, saving physicians 5 to 30 minutes of preparation per visit, while other AI systems are being tested to detect conditions such as pancreatic cancer earlier.
The hospital develops and validates AI tools through clinical trial-like testing before wider deployment, emphasizing physician oversight and ongoing performance monitoring. While Mayo says AI is improving efficiency and patient care rather than replacing clinicians, its growing use has also sparked scrutiny over patient privacy and governance, underscoring the need to balance innovation with trust in healthcare. → Read the full article here.
NEWS
What Else is Happening
Bernie Sanders Pushes AI Ownership Plan: Sanders cited a poll showing 69% support for giving the public a 50% stake in major AI companies.
Fireworks Reaches $17.5B Valuation: NVIDIA-backed AI startup raised $1.5B after topping $1B in annualized revenue as demand for open-source AI grows.
NVIDIA Expands Japan AI Push: NVIDIA launched Cosmos 3 Edge and partnered with Japanese firms to advance manufacturing and healthcare AI.
Applied Computing Raises $20M: The startup is building an AI model that combines sensor and physics data to optimize oil and gas plant operations.
NotebookLM Becomes Gemini Notebook: NotebookLM renamed to Gemini Notebook, expanding Google integration and adding a secure cloud computer.
Go Beyond Chat: A Hands-On AI Course for Working Teams
Most Teams Use AI. Few Use It Well.
Forward Future is teaming up with Hiten Shah to explore a live, hands-on cohort course for companies already using ChatGPT, Claude, Copilot, or similar tools, and ready to go further.
The course is built around your real work. Each company enrolls two or more colleagues who share a recurring workflow or business problem, and you'll spend the course actually solving it.
We're gauging interest before locking in the first cohort. The form takes about four minutes, and filling it out commits you to nothing.
FRIDAY FACT
A Single Gram of DNA Can Store 215 Million Gigabytes
In 2017, geneticists Yaniv Erlich and Dina Zielinski encoded a full computer operating system, a film, an Amazon gift card, and other files into synthetic DNA — then read all of it back with zero errors. Their method, called DNA Fountain, reached a density of 215 petabytes per gram, roughly 215 million gigabytes packed into something lighter than a grain of rice.
The reason DNA is such a good archive is the same reason it runs biology: it's compact, stable for thousands of years, and never goes obsolete the way a floppy disk or a Zip drive does. At that density, every byte humanity has ever produced would fit in a space about the size of a shipping container. The catch is speed and cost — writing DNA is still slow and expensive — which is why it's aimed at cold storage, not your laptop.
