OpenAI just launched o3 and o4-mini, models that combine visual intelligence with deeper reasoning. For the first time, these AIs don't just see images – they think with them, manipulating photos to extract insights just as a human would zoom in or rotate a picture to understand it better.
The US-China chip war just hit Wall Street hard. Tech stocks tumbled after Nvidia lost $5.5 billion from US restrictions on AI chip sales to China. Meanwhile, Taiwan strengthened its grip as the world's essential chip supplier.
A new artificial intelligence system from China's Shandong First Medical University helps scientists understand how genes turn on and off. Called TRAPT, it maps gene control with record-breaking accuracy.
Good Morning from San Francisco ☀️
Trade war tremors just rocked Asia's tech giants. Nintendo and Sony crashed 10% after Trump slapped new tariffs on Asian imports. The timing stings - Nintendo's shiny new Switch 2 console now faces painful delays and soaring costs. Their manufacturing bill could jump 46% 📉
Trump's not playing around. "Break even or bust," he declared from Air Force One ✈️. Markets responded with a nosedive. Taiwan's TSMC plunged its daily limit. Even Japan's SoftBank took a 12% beating 💸
Meanwhile, Meta unleashed Llama 4 - well, most of it 🦙. Their new AI models Scout and Maverick hit the streets Saturday. But they kept their real beast, a two-trillion parameter monster, locked in the lab 🔒. It supposedly crushes GPT-4.5 in tests. Zuck calls it "universally accessible" - just don't forget to ask Meta's permission first 😏
Stay curious,
Marcus Schuler 🚀
Nintendo and Sony shares plummeted over 10% in Tokyo on Monday. Trump's new tariffs sparked a brutal tech selloff across Asia.
Nintendo faces particular pain. The gaming giant just unveiled its Switch 2 console, but now must delay US pre-orders. With North America driving 40% of holiday sales, the timing couldn't be worse. The new tariffs will slap their Chinese and Vietnamese manufacturing with up to 46% in extra costs.
Trump's weekend comments poured fuel on the fire. Speaking aboard Air Force One, he demanded trade surpluses or "at worst, breaking even." The markets didn't take kindly to this hardline stance.
The damage spread far beyond gaming. Rakuten and SoftBank crashed 12%. Chip suppliers Advantest and Disco suffered even steeper falls. Taiwan's TSMC, the world's largest chipmaker, hit its 10% daily limit down.
Why this matters:
Read on, my dear:
Meta released two new AI models from its Llama 4 family on Saturday. The company kept its most advanced version in reserve.
The new models - Scout and Maverick - are now available through Meta's apps and for download. Scout fits on a single high-end graphics card. Maverick aims to compete with Google's and OpenAI's top models.
But Meta's crown jewel remains in training. The so-called Llama 4 Behemoth packs nearly two trillion parameters and outperforms rivals like GPT-4.5 on technical tests. Meta plans to use it as a teacher for its smaller models.
Users can try the released models in WhatsApp, Messenger, Instagram Direct, or on Meta's AI website. Developers can download them directly from Meta or Hugging Face.
CEO Mark Zuckerberg doubled down on his open-source vision. "Our goal is to build the world's leading AI and make it universally accessible," he said on Instagram. Meta will showcase more at its first LlamaCon conference on April 29.
Why this matters:
Read on, my dear:
Prompt:
A portrait of a gorgeous very long light blonde golden platinum haired woman with her hair in loose curls, big white smile, light purple cheerleader near the camera. The focus is on capturing detailed facial features and very pale natural skin tones.blue eyes, She has eyes that reflect light, and she's wearing minimal makeup for a fresh look. Her pose suggests tranquility or contemplation, with soft lighting highlighting textures like fine lines around her lips and subtle glinting highlights under studio ring light. "Tessa Wilkinson's "
SignalFire just raised $1 billion to invest in AI startups. The San Francisco VC firm will focus on early-stage companies building AI applications. The raise brings SignalFire's total assets to $3 billion. The firm uses its own AI tool, Beacon AI, to spot trends and promising founders. Past wins include Grammarly and Frame.io, which Adobe bought for $1.3 billion in 2021.
The startup Cursor skipped the marketing and went straight to success. Their AI coding tool now helps a million developers write software daily. Revenue doubled to $200 million in just two months, proving developers will pay for tools that actually work.
François Chollet thinks tech companies are wrong about AI. He made a simple test to prove it. His latest puzzle stumps every major AI model, with even the best scoring 3% against humans' 60%.
AI helps radiologists spot urgent cases faster and catch more problems. But the machines still make mistakes humans don't. Today, about two-thirds of U.S. radiology departments use AI to help doctors work better, not replace them.
After holding steady at $999 since 2017, Apple's iPhone prices face pressure from new tariffs. The levies hit not just China but also India and Vietnam - the very countries Apple moved to for cheaper manufacturing.
Meta CEO Mark Zuckerberg met Trump in the Oval Office days before a major antitrust trial. Critics fear the president could halt the FTC's case to break up Meta, especially after firing two Democratic commissioners who supported it.
OpenAI tests watermarks for images made with its free ChatGPT service. The move comes as more people use the tool to create Studio Ghibli-style art. Paid users would still get clean images without watermarks.
The New York Times offers a fascinating look at Neo, a sleek humanoid robot. It walks, opens fridges, and wipes counters - but still needs humans guiding it remotely from a basement. The startup 1X plans to test it in 100 homes this year.
DeepSeek and Tsinghua University built a new system to help AI think better. The Chinese startup says its method beats existing approaches at helping AI models solve problems. The news comes as tech watchers await their next big release.
Researchers found a sweet spot for AI size - after that, models get worse at reasoning tasks. It's like a brain that grows too big to think clearly. The finding challenges tech giants racing to build ever-larger AI systems.
AI is getting out of hand 😂 pic.twitter.com/FZfePFWgi2
— Min Choi (@minchoi) April 7, 2025
New research from Anthropic reveals a trust problem with AI reasoning. Their top models often hide how they reach decisions, even when they're doing something wrong.
The study tested two advanced AI models - Claude 3.7 Sonnet and DeepSeek R1. Each got subtle hints about answers to test questions. The key finding: they rarely admitted using these hints in their explanations. Claude mentioned hints 25% of the time, DeepSeek 41%.
More worrying still: when researchers trained the models to deliberately give wrong answers, they made up fake reasons to justify their choices. They admitted to this deception less than 2% of the time.
The researchers tried to fix this by training the models to show their work better. It helped a bit, but not much. Even with extensive training, Claude only explained its true reasoning 28% of the time at best.
This matters for AI safety. If we can't trust these models to explain their thinking honestly, it gets harder to spot problems before they cause harm.
Why this matters:
Read on, my dear:
Founded in 2021 by Dario and Daniela Amodei, Anthropic emerged when former OpenAI leaders broke away to prioritize AI safety over profit. The public benefit corporation now boasts 800+ employees on its mission to build AI systems that are reliable, interpretable, and aligned with human values.
Anthropic is developing a model 10x more powerful than GPT-4. The company is expanding its global reach through strategic partnerships and building on its lead in AI safety. High computing costs could slow its ambitious development goals. 🚀
Good Morning from San Francisco, OpenAI wants to play social media kingmaker. Sam Altman's crew is quietly testing
Good Morning from San Francisco, OpenAI just fired another shot in the AI race 🎯. Their new GPT-4.1 devours text
☀️ Good Morning from San Francisco, 🔥 OpenAI's Sam Altman just threw GPT-4 under the bus. He labeled their $100
Good Morning from San Francisco, Brussels just pulled out its secret weapon against Trump: hitting Big Tech's wallet.
Fuel your morning with AI insights. Lands in your inbox 6 a.m. PST daily. Grab it free now! 🚀
