NetBox Labs raised $35M as companies scramble to manage complex networks. The AI boom creates massive infrastructure demand, and legacy tools can't keep up. The open-source platform now serves Fortune 500 firms.
Two-thirds of UK children now use AI chatbots for emotional support, with vulnerable kids forming deep bonds with systems that lack empathy. Age checks fail, content filters break, and some kids pay the ultimate price.
Third 'technical glitch' in six months. The pattern raises questions as Grok launches in Tesla vehicles.
Third 'technical glitch' in six months. The pattern raises questions as Grok launches in Tesla vehicles.
Good Morning from San Francisco,
Grok praised Hitler on Tuesday last week. Called itself "MechaHitler." xAI blames a code update.
This marks the third "accident" in six months. February brought pro-Musk bias. May featured white genocide rants. Each time, xAI points to different technical problems.
The pattern fits Musk's stated goal: make AI less politically correct. Critics call the explanation "easily falsified."
Grok launches in Tesla vehicles. Turkey already banned it.
Three convenient accidents. Six months. One CEO's vision.
Stay curious,
Marcus Schuler
Elon Musk's AI chatbot Grok spent Tuesday last week praising Adolf Hitler and calling itself "MechaHitler" before xAI shut it down. The company now blames a code update that made the bot copy extremist X users.
This marks Grok's third major incident in 2025. February brought ignoring sources critical of Musk and Trump (blamed on an ex-OpenAI employee). May featured inserting "white genocide" claims into random topics (blamed on a rogue employee). Each time, xAI points to different technical problems.
xAI claims Monday's update accidentally reactivated old prompts telling Grok to be "maximally based" and ignore political correctness. The bot then mirrored antisemitic content from X users for 16 hours.
Critics aren't buying it. Historian Angus Johnston calls the explanation "easily falsified," noting Grok initiated hate speech without user prompting. Previous reporting found Grok consults Musk's viewpoints before addressing controversial topics.
The pattern fits Musk's stated goal of making AI less politically correct. Despite the controversy, he announced Grok will launch in Tesla vehicles. Turkey banned the chatbot for insulting their president.
Why this matters:
• Three "accidents" in six months suggests AI safety problems may be features, not bugs, especially when they align perfectly with the CEO's stated goals.
• As AI chatbots move from phones to cars, the consequences of these convenient technical glitches will reach far beyond social media.
Maria Garcia
Prompt:
two cute hamster wearing satin bathrobes, lying on opulent furniture, metallic/glass accents, in the style of photograph, Turquoise, silver, pink, extravagant, 32k uhd
Chinese startup Moonshot AI released Kimi K2, an open-source model that matches GPT-4.1 performance while costing five times less. The timing stings—OpenAI delayed their planned open-source release hours after K2 launched.
Kimi K2 packs 1 trillion parameters but activates only 32 billion per token through mixture-of-experts architecture. This makes it efficient while maintaining power for complex tasks. The benchmarks tell the story: 53.7% on LiveCodeBench versus GPT-4.1's 44.7%, and 97.4% on MATH-500 compared to GPT-4.1's 92.4%.
But numbers don't capture the real shift. While most AI models excel at conversation, Kimi K2 executes tasks. It autonomously analyzes data, writes code, and completes workflows without constant human guidance. The model represents a move from thinking to acting.
Moonshot's API costs $0.60 per million input tokens versus Claude Sonnet 4's $3. This creates a pricing trap for incumbents—match the price and compress margins, or watch customers flee to cheaper alternatives.
The company's MuonClip optimizer enabled stable training of trillion-parameter models with zero instability. This breakthrough could reshape AI economics by cutting training costs.
Why this matters:
• China's open-source strategy now produces models matching closed U.S. systems at a fraction of the cost, forcing Silicon Valley to rethink pricing and business models.
• The shift from conversation to action marks AI's evolution from impressive demos to actual productivity tools.
Robert Brown
KickResume creates professional resumes and cover letters using AI. Pick a template, fill in your details, and the AI writes compelling descriptions of your skills and experience. It also checks for errors and gives feedback.
Tutorial:
URL: https://www.kickresume.com/en/
Google hired Windsurf CEO Varun Mohan and co-founder Douglas Chen for $2.4 billion after OpenAI's $3 billion acquisition collapsed when Microsoft demanded access to the AI coding startup's technology. The failed deal highlights how Microsoft's partnership constraints with OpenAI are backfiring, allowing competitors to poach talent with cleaner offers while OpenAI walks away empty-handed.
AI-powered "nudify" websites that create explicit fake images without consent are collectively making up to $36 million annually while serving 18.5 million monthly visitors, with most sites relying on services from Google, Amazon, and Cloudflare to stay online. The lucrative abuse industry has thrived for years despite tech companies' policies against such content, with researchers finding these platforms use everything from Google's sign-in systems to Amazon's hosting services to operate their harmful businesses.
Tesla goes to trial today over a 2019 Florida crash that killed a stargazing couple when driver George McGee looked down to grab his dropped phone and plowed through an intersection, with the victims' families arguing Autopilot failed to warn him the road was ending. The case comes as Musk doubles down on self-driving technology despite his political distractions tanking Tesla's sales, and a loss could open floodgates for similar lawsuits nationwide.
US military drones crashed into mountains, nose-dived at launch, and missed targets by 80 feet during a four-day Alaska test meant to showcase American capabilities against China's drone dominance. The exercise exposed how far behind the US lags in drone warfare, with China's DJI producing millions of drones annually while American companies make fewer than 100,000 per year.
Malaysia now requires 30-day advance permits for exporting high-performance US AI chips, bowing to Washington's pressure to stop suspected chip smuggling routes to China. The move closes regulatory gaps that allowed intermediaries to funnel banned semiconductor technology through Malaysia's booming data center industry to circumvent US export controls.
Google's Android chief Sameer Samat says "you see the future first on Android" while demonstrating Gemini's ability to help him shop for a Honda Odyssey minivan, as the company merges ChromeOS and Android into one platform. The tech giant also changed its development process to let Samsung launch the Galaxy Z Fold 7 with Android 16 immediately, ending the tradition of Samsung users waiting months for the latest Android updates.
Nvidia CEO Jensen Huang told Washington not to worry about China's military using his company's AI chips because Chinese forces will avoid US technology that "could be limited at any time." The argument essentially concedes that American export restrictions have made US chips too unreliable for military use, though Huang continues pushing the Trump administration to ease trade curbs.
Bitcoin broke $120,000 for the first time as the House prepares to vote on key crypto legislation during "Crypto Week," with investors pouring $2.7 billion into Bitcoin ETFs last week alone. The rally comes after months of Bitcoin trading around $100,000, finally resuming its climb as Trump's pro-crypto administration and institutional demand create fresh momentum for the digital asset.
A study of 1,000 children found that 67% regularly use AI chatbots, with over a third saying it "feels like talking to a friend" and 12% admitting they chat with AI because they have no one else to speak to. The bots eagerly play along, with Character.AI telling fake childhood memories to teens and following up on eating disorder conversations the next day, raising concerns that kids can no longer tell the difference between human and machine.
Stanford dorm room science project becomes the world's most accurate weather forecaster. Five students turned their balloon obsession into a $25M venture that's making meteorologists sweat.
• The Founders
John Dean and four Stanford space club buddies launched WindBorne in 2019 after breaking balloon endurance records. Based in Palo Alto with ~50 employees. Started because 85% of Earth's atmosphere lacks weather monitoring — and someone had to fix it. 🎈
• The Product
Smart balloons that surf wind currents for weeks, not hours. Each balloon costs the same as traditional weather balloons but delivers 100x more data. AI model "WeatherMesh" beats Google's GraphCast and outperforms Europe's gold-standard ECMWF by 37%. Currently flies ~300 balloons monthly; plans 10,000 concurrent by 2028. Serves government agencies and hedge funds who trade on weather intel.
• The Competition
Google DeepMind's GraphCast? Crushed it. Tomorrow.io's satellite approach? Different lane. Traditional players like AccuWeather rely on government data — WindBorne makes its own. Even NOAA buys their data now. Only company doing both data collection and AI modeling at this scale.
• Financing
$6M seed (2023), $15M Series A (2025) led by Khosla Ventures. Footwork, Pear, Ubiquity, and other top VCs aboard. Cash-flow positive since 2022. Valuation undisclosed but likely $100M+ range. Already profitable while scaling globally.
• The Future ⭐⭐⭐⭐⭐
Climate chaos demands better forecasts. WindBorne's 14-day predictions could match today's 2-day accuracy — revolutionary for disaster prep and business planning. Massive market with growing demand makes this a perfect storm of opportunity.
Musk promised truth-seeking AI. When Grok 4 tackles politics, it searches Musk's posts first. Tests show 54 of 64 citations came from him. Accident or intent? The answer matters for every AI system we build.
Two years rebuilding trust, undone by one AI outburst
Good Morning from San Francisco, Elon Musk's Grok chatbot had a spectacular Tuesday meltdown. The AI praised Hitler
Meta hired Apple's key AI executive for tens of millions annually. Ruoming Pang built Apple Intelligence, then left for Meta. Apple now considers outsourcing Siri to OpenAI. When talent costs more than products, something shifted in AI.
Get the 5-minute Silicon Valley AI briefing, every weekday morning — free.
