Sahil Lavingia joined Elon Musk's government efficiency team with Silicon Valley confidence. Fifty-five days later, he got fired for telling reporters what he found inside. His account reveals who really ran DOGE—and why the VA surprised him.
Palmer Luckey got fired from Facebook for backing Trump. Now Meta needs his defense company to win a $22 billion military contract. The reunion changes everything.
Palmer Luckey got kicked out of Facebook for backing Trump. Now Meta wants him back. 🤝
Luckey's defense company Anduril just snatched a $22 billion military VR contract from Microsoft. Meta suddenly needs what their former outcast built. 💰
China's tech giants are dumping Nvidia. Alibaba, Tencent, and Baidu stockpiled American chips before export restrictions hit. Those stockpiles run dry by early next year. 🇨🇳
Trump's new rules blocked Nvidia's watered-down H20 chip. Chinese companies now test Huawei alternatives despite three-month development delays. Nvidia warned investors of $8 billion in lost revenue this quarter. 📉
The hunter became the hunted. 🎯
Stay curious,
Marcus Schuler
From Silicon Valley Exile to Defense Contractor Gold
Palmer Luckey got fired from Facebook for backing Trump in 2017. Now Meta partners with his defense company to build military VR headsets worth millions.
The reunion centers on EagleEye headsets that detect drones miles away and control AI weapons systems. Luckey's company Anduril leads a $22 billion Army program after wresting control from Microsoft, which spent years failing to deliver working military headsets.
Meta brings VR hardware expertise and AI models. Anduril contributes battlefield software and military contracts knowledge. Together they bid for a $100 million Army prototype contract, though the partnership continues regardless of that outcome.
The collaboration marks Meta's biggest defense push. The company recently opened its AI models for military use and hired former Pentagon officials. CEO Mark Zuckerberg says the technology helps American soldiers "protect our interests at home and abroad."
Luckey founded Oculus at 15 and sold it to Facebook for $2 billion at 21. His 2016 donation to an anti-Hillary Clinton group triggered internal pressure to remove him. Facebook fired him despite protests the donation was personal, not political.
He used Oculus money to start Anduril in 2017, betting Trump's return would create defense opportunities. Microsoft originally won the $22 billion military headset contract in 2018 but lost program leadership in February after years of technical failures. Software updates that took Microsoft 180 days now reach soldiers in 18 hours under Anduril.
The partnership reflects Silicon Valley's shift from anti-war sentiment to competing for defense contracts. Other tech companies follow suit as Trump's administration pushes military adoption of commercial technology. For Meta, defense work offers new revenue beyond advertising after spending $13.7 billion on Reality Labs last year.
China invests heavily in military AI while traditional American defense contractors move slowly. Startups like Anduril promise faster innovation by adapting commercial technology for military use. The approach could save money while accelerating development.
Eight years after his firing, Luckey leads this partnership. Meta needs Anduril more than Anduril needs Meta. The former outcast now controls access to military VR contracts his old employer wants.
Why this matters:
Luckey's comeback shows how quickly Silicon Valley's political winds shifted—what got him fired in 2017 now makes him valuable in 2025.
Military contracts represent tech's new gold rush as companies seek revenue beyond traditional streams while facing regulatory pressure.
Prompt: a black girl wearing white headphones close up, straight on, can see whole face
The Great Chip Divorce: How Trade Wars Are Reshaping AI Hardware
China's tech giants are breaking up with Nvidia. Alibaba, Tencent, and Baidu have started testing domestic chip alternatives as their stockpiles of American processors dwindle. The companies expect their current Nvidia inventory to run out by early next year.
The urgency stems from Trump's latest export restrictions, which blocked sales of Nvidia's H20 chip—a watered-down version designed to comply with earlier Biden-era rules. New chip orders take three to six months to ship, and Nvidia hasn't announced when it might offer a compliant replacement.
Baidu's AI chief Shen Dou told analysts his company could choose from various chip options to replace Nvidia's hardware. Alibaba and Tencent made similar statements about exploring alternatives, though switching carries costs. One tech executive estimated the migration to Huawei's chips would cause three months of development delays.
Meanwhile, Nvidia fought back with an unusual earnings strategy. The company led its investor presentation by highlighting the damage from export controls—an estimated $8 billion revenue loss this quarter. That figure exceeded even pessimistic analyst projections.
CEO Jensen Huang argued that Washington's strategy has backfired. He pointed to China's DeepSeek AI model and claimed Huawei's latest chips match Nvidia's previous-generation hardware. His message: restrictions have failed to stop China's AI development and only hurt American companies.
The approach worked with investors, who pushed Nvidia's stock higher. But Washington wasn't listening. The day after Nvidia's results, regulators extended chip restrictions and forced another tech company, Synopsys, to suspend its financial guidance.
State-owned Chinese companies already buy Huawei's Ascend chips in large numbers. Now private tech firms are expected to join them, despite limited supply from domestic manufacturers.
Why this matters:
Export controls meant to slow China's AI progress are instead accelerating its push for chip independence—exactly the opposite of their intended effect.
Nvidia's public lobbying campaign shows how quickly geopolitics can reshape entire industries, even when the economics favor free trade.
I want you to act as a YouTube video analyst. After I share a video link or transcript, provide a comprehensive analysis in approximately 100 words using a clear, engaging paragraph format.
Requirements:
Extract and chronologically organize the creator's key ideas and main arguments
Include significant quotes with timestamps when available
Highlight any future predictions or forward-looking statements
Focus solely on the video's core messages—no tangential information
Write in an engaging, accessible style that flows naturally
Format: Single cohesive paragraph that tells the story of the video's content chronologically.
{Insert link or transcript here}
German AI Startup Takes on Google and OpenAI with Image Editing Breakthrough
Black Forest Labs from Freiburg has solved a problem that others struggle with. The German startup showed with FLUX.1 Kontext how you create and edit images in one workflow.
The new AI does both at once. Enter text, upload image, done. No complex workflows, no trial and error. The system understands what users want and delivers.
The secret lies in the technology. Instead of using diffusion like everyone else, Black Forest Labs uses flow matching. That sounds abstract but brings concrete benefits. Editing runs up to eight times faster than competitors.
Two versions are ready. Kontext Pro for iterative workflows, Kontext Max for maximum speed. Both maintain characters and styles across multiple editing steps. A detail that often fails with others.
The timing fits. Google just launched Imagen 4, OpenAI upgrades ChatGPT. The market bubbles. Investors noticed and talk about $100 million at a billion valuation.
The founders come from Stability AI, the team behind Stable Diffusion. They know how to make image generation work. FLUX.1 Dev was downloaded 30 million times already.
Black Forest Labs also launched Playground, a browser demo for business users. Teams can test use cases and show stakeholders what the technology does before committing.
Why this matters:
German AI company shows Europe can compete in generative AI – while others only talk about regulation
Image editing becomes as simple as sending a WhatsApp message – no Photoshop skills needed
Elon Musk used drugs far more intensely than previously known while becoming one of Trump's closest allies last year. He told people his ketamine use was affecting his bladder and traveled with a daily pill box containing about 20 medications, including ones marked as Adderall. Meanwhile, he juggled overlapping romantic relationships and custody battles over his 14 known children, including a secret relationship that produced another child whose mother he offered $15 million to stay quiet.
DeepSeek Quietly Drops AI Update That Beats Google Models
DeepSeek released an updated reasoning model Thursday without fanfare that now ranks just behind OpenAI's best on coding tests. The R1-0528 update jumped math accuracy from 70% to 87.5% and coding scores from 63.5% to 73.3% by using twice as many thinking tokens per question. The Chinese company also created an 8-billion parameter version that matches models 30 times larger, proving smart algorithms can beat expensive hardware.
Meta Cuts Content Removals by One-Third After Free Speech Push
Perplexity Launches AI Tool That Builds Reports and Spreadsheets
Perplexity launched Labs Thursday, a new tool for Pro subscribers that creates reports, spreadsheets, and dashboards using AI. The feature takes about 10 minutes to complete tasks, using web search and code execution to build interactive apps and analyze data. Perplexity continues pushing beyond search as it reportedly seeks $1 billion in funding at an $18 billion valuation.
Hugging Face Drops Humanoid Robots for Under $3,000
Hugging Face announced two humanoid robots this week that won't break your budget or hide their code. The HopeJR walks around with 66 degrees of freedom for roughly $3,000, while the desktop Reachy Mini talks and listens for under $300.
This German startup emerged from the Stable Diffusion team to challenge Silicon Valley's generative AI dominance. Their FLUX models already outperform OpenAI and Midjourney while maintaining an open-source edge.
The Founders Founded in 2024 by Robin Rombach and nine former Stability AI researchers in Freiburg, Germany. The team includes the original architects of Stable Diffusion from University of Munich. Company employs roughly a dozen core team members who left Stability AI amid leadership turmoil to build next-generation image models on their own terms.
The Product FLUX family of text-to-image models built on cutting-edge "flow matching" architecture. Core strengths: photorealistic output, flawless human anatomy rendering, legible text generation, and unified image creation/editing in one framework. Offers both open-source versions (30M+ downloads) and commercial API. Generated 700M+ images via API by late 2024. Speeds up to 8x faster than competitors with 4-megapixel outputs in 3-5 seconds.
The Competition Direct rivals include OpenAI's DALL-E, Midjourney, and former employer Stability AI. FLUX already leads usage on platforms like Hugging Face and Poe, displacing established players. Adobe, Google, and Microsoft represent big tech threats, but their closed systems leave room for open alternatives. Video generation battleground emerging with Runway ML, Pika Labs, and OpenAI's Sora.
Financing $31M seed round (August 2024) led by Andreessen Horowitz, General Catalyst, plus angels like Oculus co-founder Brendan Iribe. Reported $200M Series A talks at $1B valuation would create Europe's fastest AI unicorn.
The Future ⭐⭐⭐⭐⭐ Plans U.S. expansion and text-to-video model codenamed "SOTA." Strong technical foundation, proven team, and open-source strategy position them perfectly for the next AI wave. Europe finally has a generative AI champion that can compete with Silicon Valley giants. 🚀
