- Yaro on AI and Tech Trends
- Posts
- 🛂Can the United States be the gatekeeper of global AI infrastructure?
🛂Can the United States be the gatekeeper of global AI infrastructure?
Kalshi and Polymarket Are Flying High. Can They Outrun the Regulators?
Check your email (including spam) for our email with the subject: “Complimentary subscription to Yaro on AI and Tech Trends | Your Top AI Newsletter.” Please reply to it with a quick “Hello” to include you in our Substack newsletter as we move towards that platform. Thanks.
Get in Front of 50k Tech Leaders: Grow With Us
I finally downloaded Kalshi this weekend, and got a crash course in what a real dopamine rush feels like. Social media has nothing on these prediction platforms. We break down how Kalshi and Polymarket are performing and the regulatory headwinds they’re facing, among other issues. We also dig into the top 100 consumer AI apps for March 2026, and the question that’s becoming impossible to ignore: Can the US actually control who gets to build AI infrastructure, or is that ship already sailing?
Let’s get into it. Stay curious.
The US Moves to Control Who Gets to Build AI Infrastructure
Can the United States be the de facto gatekeeper of global AI infrastructure?
đź§° AI Tools - Ethical and Conversational AI
📚Learning Corner - Anthropic Courses
The top 100 consumer AI apps
Kalshi and Polymarket Are Flying High. Can They Outrun the Regulators?
đź“° AI News and Trends
Claude Opus 4.6 found more high-severity bugs in Firefox over two weeks than the rest of the world typically reports in two months.
OpenAI robotics and hardware executive Caitlin Kalinowski resigned following the company's reported partnership with the Pentagon.
Microsoft, Google, Amazon to Keep Selling Anthropic to Customers Other than the Pentagon
Oracle and OpenAI will not expand their data center footprint at Abilene, Texas.
Cursor to prioritize contracts with large enterprises in an effort to remain the top Coding agent after competition from Anthropic code and OpenAI Codex gets tougher.
Other Tech News
The controversial gaming platform, Roblox, has grown quite fast since 2004, minted several teen millionaires, and paid out $1.5 billion to game creators in 2025.
Some people are using a second dumb phone to minimize distractions and are gaining momentum.
Grammarly is using our identities without permission.
China could see widespread use of brain-computer tech in 3-5 years. Policies such as insurance integration and national standards aim to close the “huge” gap between scientific research, industry, and clinical applications
The US Moves to Control Who Gets to Build AI Infrastructure
The US Commerce Department is drafting regulations that would require U.S. approval for nearly all global exports of AI chips from companies such as Nvidia and AMD. Under the proposed framework, buyers of smaller quantities (under 1,000 GPUs) would undergo a simple review process, while larger data center projects would require pre-clearance and may be subject to conditions such as business disclosures or government site visits. The biggest deployments, over 200K GPUs, would require host governments to make security commitments and matching investments in American AI.
The framework is still in draft form and faces pushback even within the Trump administration, which has distanced itself from Biden’s prior “AI diffusion rule” while not yet finalizing a replacement.
Can the United States be the de facto gatekeeper of global AI infrastructure?
Yes, the US can, but it’s not that simple:
Nvidia dominates the AI chip market by a wide margin, and there’s currently no comparable alternative. If you want the best hardware, you essentially have to go through American companies.
The US already demonstrated this leverage with the UAE and Saudi Arabia deals, countries accepted conditions that were dollar-for-dollar investment matching, restrictions on serving Chinese firms, in exchange for chip access.
Washington can also threaten that using Huawei chips could violate American trade restrictions, effectively closing off the only real alternative.
And NO, because:
China is actively developing its own chip ecosystem through Huawei and others. The chips are less powerful today, but that gap could narrow over time, and the pressure of US restrictions accelerates China’s motivation to close it.
Foreign governments are already chafing at the idea of subjecting their AI futures to Washington’s decisions. Europe, India, and others are pursuing “AI sovereignty” strategies partly in response to this dynamic.
The rule’s effectiveness depends entirely on execution. Slow licensing, political strings, or unpredictable conditions tied to trade negotiations (as the Bernstein analyst warned) could push countries to seek alternatives or build domestic capacity.
Even within the Trump administration, there’s disagreement; a White House official already said the draft doesn’t reflect the president’s direction.
The deeper tension is that the US wants two things simultaneously: to be the world’s AI supplier and to use that position as geopolitical leverage. Those goals can conflict. Push too hard on conditions and countries diversify away; be too permissive, and you lose the leverage entirely. Whether Washington can thread that needle is genuinely unclear.
📚Learning Corner
All Anthropic Educational Courses on GitHub.
The Top 100 Consumer AI apps

Courtesy of A16Z
The latest ranking of the top 100 consumer AI apps paints a picture of a maturing market. For the first time, the list now includes mainstream tools like CapCut, Canva, and Notion, where AI has quietly become central to the core experience rather than a bolt-on feature.
ChatGPT still dominates, with 900 million weekly active users, but competitors are closing in on specific fronts. Claude’s paid subscribers grew over 200% year-over-year, Gemini had a breakout year on the creative side, and roughly 20% of ChatGPT users now also use Gemini in any given week. The platforms are diverging in strategy. OpenAI is chasing the consumer super-app (shopping, travel, health), while Anthropic is doubling down on power users, developers, knowledge workers, and professionals.
Geographically, the market is split into three. Western tools, a Chinese ecosystem anchored by DeepSeek and Doubao, and an emerging Russian pole built around Yandex and Sber. The US ranks just 20th in per-capita AI adoption; Singapore, UAE, and Hong Kong lead.

Courtesy of A16Z
Creative tools have shifted from image-dominated to a broader mix of video, music, and voice, with Chinese video models leading on quality and standalone image generators losing ground to bundled features inside ChatGPT and Gemini.
Agents are the new frontier. Vibe coding tools like Lovable and Replit are holding their audiences, while horizontal agents, OpenClaw (acquired by OpenAI), Manus (acquired by Meta for ~$2B), and Genspark, are starting to let everyday users hand off complex tasks entirely. The next six months will reveal whether consumers consolidate around one AI agent or keep juggling several.
Kalshi and Polymarket Are Flying High. Can They Outrun the Regulators?

I’m probably late to the party, but I finally downloaded Kalshi this weekend after a conversation with some cousins-in-law in their mid-20s, and it was a good reminder of why it pays to listen to younger people when trying to understand where markets are heading.
Opening the app for the first time was a jolt. I thought social media was addictive. I hadn’t seen anything yet. The range of what you can trade on is staggering: sports, politics, the economy, the weather. Some markets resolve in seconds. Others play out over months. The company’s CEO insists it’s not a betting platform, but when you’re staring at a screen full of live odds, and your finger is hovering over a position, that distinction feels academic.
I had to actively stop myself from putting money in. That experience alone tells you everything you need to know about why regulators are paying close attention. It’s not just the financial implications of mainstream prediction markets; it’s the design. These apps are built to pull you back in, and they’re very good at it. The dopamine loop is real, and it’s been refined. We may look back at social media as the warm-up act.
Kalshi and Polymarket, the two biggest players in the space, are both in early talks with investors about raising new rounds at valuations near $20 billion each, roughly double where they were just a few months ago.
Kalshi, which is already live in the US and recently crossed a $1.5 billion annualized revenue run rate, was last valued at $11 billion after raising $1 billion in December from Sequoia and Paradigm. Polymarket, still blocked for US users but planning a domestic launch this year, was last valued at $9 billion following a deal with NYSE parent Intercontinental Exchange.
The growth is real, but so is the turbulence. Both platforms are facing a bipartisan bill in Congress that would ban markets on war, geopolitical events, and sports. They’re also under fire for aggressive college marketing tactics, including a case where Polymarket gave a fraternity cash in exchange for recruiting new users, and a wave of suspicious trades around Jeff Bezos’ Super Bowl whereabouts, placed by members of his stepson’s fraternity.
Prediction markets have genuine utility as forecasting tools and are growing fast, but their push into sensitive topics and college campuses is drawing exactly the kind of regulatory attention that could reshape the business before it fully matures.
đź§° AI Tools of The Day
Ethical and Conversational AI
Reflection - Private AI coach helping users find clarity, improve mental health, and unlock personal growth through journaling.
Wysa — An AI mental health companion that uses CBT, DBT, and mindfulness to help users manage anxiety, depression, and stress, with an optional human coaching upgrade.
Pi (by Inflection) — A conversational AI built for personal, empathetic dialogue. It sits between ChatGPT and a therapist. No structured output, no frameworks, no decision journaling — just conversation.
Rocky - An AI coaching platform that engages users in daily reflections, goal-setting, and learning new skills. It uses the GROW coaching model (Goal, Reality, Obstacles, Way Forward) oriented toward goal-setting and problem-solving.
Youper - Bills itself as an “emotional health assistant” using CBT and ACT to guide daily check-ins, mood tracking, and personalized exercises targeting anxiety, depression, and self-confidence.
Reply