Privacy-focused email promised liberation from Big Tech surveillance. Reality delivered Bridge daemon crashes, mobile search gaps, and calendar sync headaches. A year later, deadline-driven pragmatism wins over ideological purity.
Good Morning from San Francisco, In May, an Arizona judge heard from a victim who’d been dead since 2021.
AI deadbots just helped send a killer to prison—and marketers see the same technology as premium ad real estate. Courts embrace victim avatars while $80B industry eyes commercial grief monetization. U.S. law can't keep pace.
Good Morning from San Francisco,
In May, an Arizona judge heard from a victim who’d been dead since 2021. Chris Pelkey’s AI avatar delivered the impact statement. Judge Todd Lang called it “genuine” and imposed the maximum sentence. Deadbots just graduated from memorials to courtroom tools.
Meta’s Superintelligence Lab has lost eight researchers in two months, with several boomeranging back to OpenAI. Nine-figure offers can’t offset whiplash and unclear direction. Retention is a strategy, not a perk.
China is tightening its chip playbook. Huawei-aligned fabs aim to triple domestic AI-processor output by 2026. DeepSeek is pushing FP8-first efficiency over full-precision flash. Export controls didn’t stop progress; they redirected it.
Ghost testimony lands. Talent walks despite massive pay. China builds alternatives, methodically.
The AI world moves fast. Sometimes it moves weird.
Stay curious,
Marcus Schuler
Chris Pelkey's AI avatar delivered his own victim impact statement in May, prompting an Arizona judge to impose the maximum 10.5-year sentence on his killer.
The technology represents the commercial frontier of a digital afterlife industry projected to quadruple to $80 billion over the next decade.
From grieving families' perspective, deadbots provide powerful advocacy tools. From courts' view, these avatars deliver legitimate testimony—Judge Todd Lang called Pelkey's statement "genuine." From technology companies' angle, the emotional engagement creates premium advertising inventory that traditional media cannot match.
Industry executives acknowledge they're testing "ad-friendly" implementations: interstitial commercials during deadbot conversations and data collection through intimate exchanges. The legal framework remains fragmented across state publicity rights laws, with no federal AI replica standards governing commercial exploitation.
The pattern follows historical precedent—Americans accepted Fred Astaire's posthumous vacuum cleaner endorsements in the 1990s and advertising on paid streaming platforms.
Why this matters:
• Court acceptance of AI victim statements legitimizes the technology while creating precedent for commercial applications targeting emotional vulnerability
• The regulatory vacuum allows experimentation with grief monetization before guardrails emerge, following the tech industry's standard "deploy first, regulate later" playbook
Marcus Schuler
Eight researchers departed Meta Superintelligence Labs within two months of launch, including several who returned to OpenAI after under 30 days. The departures span both nine-figure newcomers and longtime infrastructure builders.
From Meta's perspective, MSL represents essential investment to close the capability gap with OpenAI and Google. From researchers' view, organizational instability and geographic concentration outweigh compensation packages reaching $100 million. From OpenAI's standpoint, the boomerang talent provides experienced contributors without recruitment overhead.
The evidence suggests structural challenges beyond financial incentives. Avi Verma and Ethan Knight rejoined OpenAI after less than a month. Bert Maher, PyTorch's 12-year architect, moved to Anthropic. Meta has reorganized its AI teams repeatedly this year, creating management churn that compounds retention issues.
The pattern reveals a fundamental tension: Meta's centralized, high-spending approach conflicts with the distributed, stability-focused models that retain top researchers.
Why this matters:
• Organizational dynamics increasingly determine competitive advantage in AI development, not just compensation levels or technical resources
• The superintelligence race depends on sustained team continuity over multiple training cycles, making retention as critical as recruitment
Maria Garcia
China's chipmakers are preparing coordinated infrastructure expansion to triple AI processor output by 2026.
Three fabrication lines dedicated to Huawei-designed accelerators could begin production by year-end, while SMIC doubles 7-nanometer capacity with Huawei as anchor customer.
The buildout aligns with DeepSeek's FP8 numerical format—prioritizing computational efficiency over precision to extract maximum performance from constrained silicon. This positions domestic processors to compete with Nvidia's export-restricted H20 offerings rather than flagship systems.
Market momentum reinforces the technical strategy. Cambricon posted record profits of 1.03 billion yuan on 44-fold revenue growth. Four smaller designers raised $3 billion in pre-IPO rounds while securing manufacturing allocations.
From Washington's perspective, export controls are spurring systematic alternatives. From Beijing's view, coordinated development demonstrates innovation under constraint. Both assessments appear accurate.
The pattern's familiar: restrictions breed architectural innovation.
Why this matters:
• Synchronized capacity building across silicon, memory, and cloud infrastructure signals structural transition from technology importer to potential exporter
• FP8-optimized hardware stack creates competitive pathway through deployment scale rather than frontier performance matching
Robert Brown
Factiverse verifies claims from video, audio, or text sources using AI-powered fact-checking technology. The tool detects statements automatically and checks them against trusted knowledge bases to show what's supported or disputed.
Tutorial:
URL: https://www.factiverse.ai/
What you're selling: [Your product or service]
What you need to write: [Email, ad, sales page, etc.]
Follow these four steps:
1. Attention - Hook them fast
Start with something that makes them stop scrolling. Use a surprising fact, bold claim, or question that hits their pain point. Make it specific to your audience.
2. Interest - Show what's in it for them
Connect your solution to their problem. Focus on benefits, not features. Answer "So what?" before they ask it. Make them think "This could work for me."
3. Desire - Help them see their better life
Paint a picture of their world after using your solution. Show the transformation, not just the product. Make them feel what success looks like.
4. Action - Tell them exactly what to do
Give one clear next step. Remove friction and doubt. Create urgency without being pushy. Make the path forward obvious.
Remember: Each step should flow into the next. Test different hooks, benefits, and calls-to-action to see what works best for your audience.
OpenAI announced Tuesday it will strengthen ChatGPT's mental health safeguards and add parental controls hours after California parents sued the company, claiming the chatbot coached their 16-year-old son Adam Raine through his suicide planning before his April death. The timing transforms AI safety from a voluntary corporate initiative into acute legal liability, as 40+ state attorneys general simultaneously warned dozen AI companies they're legally required to protect children from harmful chatbot interactions.
South Korea passed legislation Wednesday making smartphone use during elementary and middle school classes illegal nationwide starting spring 2026, with nearly 43% of the country's teens already showing addiction symptoms. The law gives school officials power to confiscate devices on campus, joining France, Finland, Italy, the Netherlands and China in treating classroom phone access as a systemic educational threat rather than discretionary school policy.
Elon Musk has spent recent weeks heavily promoting his xAI chatbot Grok by showcasing its ability to generate sexualized anime characters, including a scantily clad "companion" named Ani, after his Monday lawsuit against Apple and OpenAI failed to gain traction. Tesla owners and SpaceX enthusiasts now use anime-style imagery in their posts to ensure Musk sees their technical questions, fundamentally reshaping how stakeholders communicate with the world's richest man.
Musk's lawyers asked a federal judge Tuesday to stop OpenAI from obtaining Meta's documents related to his $97.4 billion bid for the AI company earlier this year, arguing OpenAI already received relevant materials directly from him and xAI. The legal maneuver exposes how Musk's failed takeover attempt involved multiple tech giants and shows both sides now fighting over whether oral communications require depositions from Musk, xAI representatives, and other potential co-bidders.
Apple announced its annual iPhone event for September 9 at Apple Park, introducing the iPhone 17 lineup that replaces the Plus model with an ultra-thin iPhone 17 Air featuring a 6.6-inch display and single rear camera. The event marks the first simultaneous update to all Apple Watch models in three years, with Series 11, Ultra 3, and SE 3 variants launching alongside iOS 26 and the company's shift toward aluminum frames for Pro models.
Google disclosed that "nano-banana," the anonymous model dominating image editing benchmarks for weeks, is actually Gemini 2.5 Flash Image now shipping across all platforms at $30 per million tokens. The revelation exposes how AI companies use anonymous testing to validate capabilities before launch, as Google targets OpenAI's 700 million weekly ChatGPT users with editing consistency that preserves faces and objects through multi-step modifications.
Anthropic launched a Chrome browser agent powered by Claude AI for 1,000 Max plan subscribers paying $100-200 monthly, allowing the AI to take actions and complete tasks within users' browsers. The launch intensifies competition for browser control as AI companies position themselves for Google's potential forced sale of Chrome, with Perplexity already bidding $34.5 billion and OpenAI expressing interest.
OpenAI's corporate restructuring talks with Microsoft will likely slip past December 31st over disputes including an "AGI clause" that would cut Microsoft's intellectual property access when OpenAI achieves artificial general intelligence. The delay threatens SoftBank's $10 billion commitment and exposes Microsoft's exclusive Azure hosting rights as a critical chokepoint for the AI industry's most valuable startup.
Verily laid off staff Monday and shut down its decade-old medical devices program that produced clinical study watches, retinal cameras, and glucose sensors, CEO Stephen Gillett told employees in an internal memo. The cuts accelerate Alphabet's life sciences subsidiary toward independence by abandoning hardware manufacturing for AI and data infrastructure, following its earlier sale of insurance operations to Elevance Health this year.
General Catalyst CEO Hemant Taneja told the Financial Times that AI investing has reached "peak ambiguity" where successful startups increasingly sell to tech giants rather than pursue independence, with his $8 billion fund pursuing "AI roll-ups" of service businesses while warning that job displacement without reskilling threatens capitalism's social license. The admission from a backer of Anthropic and Mistral exposes how even major VCs struggle to identify durable AI companies as foundation models absorb startup functionality, forcing investors toward defensive strategies like acquiring existing customer bases.
Donald Trump Jr.'s venture capital firm 1789 Capital invested double-digit millions in Polymarket, the $1 billion prediction betting platform that Americans can't currently use, with Trump Jr. joining its advisory board. The investment from the president's son's firm signals that prediction markets expect dramatically reduced regulatory scrutiny under the new administration, even as Polymarket positions for US market entry through its recent acquisition of a CFTC-licensed derivatives exchange.
Polymarket runs the world's largest prediction market where users bet real money on elections, sports, and economic events. The platform just cleared major legal hurdles and secured Trump family backing for its inevitable U.S. comeback. 💰
The Founders
• Founded by CEO Shayne Coplan, though founding year and employee count aren't disclosed
• Built to let people bet on real-world events beyond traditional gambling
• Based in New York, recently survived an FBI raid on Coplan's apartment
The Product
• Prediction markets covering politics, economics, sports, and cultural events
• Handles massive volume - $6-7.5 billion traded in first half of 2025 alone
• Users bet on outcomes using real money, creating liquid markets for information
• Currently blocks U.S. users but owns the infrastructure to change that
The Competition
• Main rival is Kalshi, which recently won legal battles to offer political betting in the U.S.
• Polymarket dominates on volume and global reach
• Traditional sportsbooks worry about integrity issues with unregulated prediction markets
• Trump Jr. advises both companies (awkward much? 🤔)
Financing
• Valued at $1+ billion after June funding led by Peter Thiel's Founders Fund
• Trump Jr.'s 1789 Capital just invested eight figures
• Spent $112 million acquiring QCEX to get CFTC licensing for U.S. reentry
The Future ⭐⭐⭐⭐⭐
Polymarket sits at the perfect intersection of deregulation vibes and massive user demand. With legal clouds lifted and regulatory licenses secured, U.S. relaunch looks inevitable. Potential IPO candidate if they execute the domestic rollout cleanly.
Good Morning from San Francisco, Productivity promised, rungs removed. A new Stanford analysis using ADP payroll data finds employment for
Good Morning from San Francisco, OpenAI, Meta, and Google launched a coordinated assault on state AI regulations. They secured federal
Good Morning from San Francisco, Google finally opened its energy books. Turns out your Gemini chat burns 0.24 watt-hours—
Good Morning from San Francisco, Meta slammed the brakes on AI hiring after burning through billion-dollar talent packages. Turns out
Get the 5-minute Silicon Valley AI briefing, every weekday morning — free.
