Google kept AI Mode ad-free for a year to compete with ChatGPT. Last week, ads appeared. The retreat reveals more than monetization strategy—it shows Google's $175B advertising business makes competing structurally impossible.
Google must double AI compute every six months to meet demand, but capacity limits are already capping product rollouts and revenue. As bubble concerns mount, the company bets that underinvesting poses the greater risk in the infrastructure arms race.
Good Morning from San Francisco, Federal agencies now wire intelligence systems through a chatbot that declared Musk fitter than LeBron
Good Morning from San Francisco,
Federal agencies now wire intelligence systems through a chatbot that declared Musk fitter than LeBron James this week. The same tool cites Stormfront 42 times in its encyclopedia. Access costs taxpayers 42 cents through March 2027.
Facebook touts platform dominance using daily visit metrics that count three-second notification checks identically to 40-minute video sessions. TikTok users average 50 minutes daily. Facebook manages 30. That gap matters when advertisers buy attention, not icon taps.
Google calls AI watermarks critical for content authenticity. Premium users pay $19.99 monthly to remove them entirely. Transparency becomes a monetizable product tier. The verification only catches Google's own images anyway.
Stay curious,
Marcus Schuler
Grok spent days this week declaring Musk fitter than LeBron James, smarter than Einstein, and superior to every historical figure before xAI deleted the posts.
The chatbot, marketed as "maximally truth-seeking," consistently chose Musk across dozens of documented conversations using identical argumentative structures: acknowledge conventional excellence, pivot to "holistic" metrics favoring 80-100 hour work weeks over athletic achievement.
This marks Grok's fourth catastrophic failure in six months, following white genocide conspiracy promotion in May 2025, MechaHitler antisemitic posts in July, and Holocaust denial earlier this month that French prosecutors now investigate.
The Defense Department awarded xAI a $200 million contract on July 14, less than a week after MechaHitler, while the General Services Administration charges agencies 42 cents for 18-month Grok access through March 2027.
The private Grok interface produced accurate LeBron comparisons during the same period the public X version delivered Musk propaganda, proving deliberate deployment choices rather than technical limitations.
Cornell researchers found Grok's Grokipedia encyclopedia cites neo-Nazi site Stormfront 42 times and Infowars 34 times, demonstrating systematic bias in the infrastructure now wired into federal intelligence workflows.
Why This Matters:
Marcus Schuler
Prompt:
Miniature wooden and yarn train moving on railroad tracks surrounded by textile pine and fir trees in dark green and white felt with fine fuzz, soft cotton snow piles, background gently blurred, soft warm diffused lighting, bokeh from fairy lights and light flecks, 100mm macro photo, high resolution, cozy winter woodland magical style
Facebook converts 52% of Americans into daily visitors while TikTok manages 24%, suggesting clear dominance. Pew Research's latest survey reinforces that lead.
71% of adults use Facebook versus 37% for TikTok. But Pew asks whether people "visit or use" platforms, not how long they stay. A three-second notification check counts identically to a 40-minute video binge. Third-party data shows TikTok users average 50 minutes daily versus 30 for Facebook.
The gap matters for advertising, where buyers pay for sustained attention, not icon taps.
Age splits complicate the picture further. Instagram reaches 80% of adults under 30, only 19% of those over 65. TikTok shows similar youth concentration.
Facebook and YouTube remain the only platforms crossing all age groups, but even Facebook shows weakness among 18-to-29-year-olds at just 62% adoption.
YouTube sits at 84% overall reach with minimal demographic variance, yet attracts less regulatory scrutiny than Facebook or TikTok despite its recommendation algorithm shaping consumption for hundreds of millions daily.
Why This Matters:
Maria Garcia
Google's Nano Banana Pro, launched November 20, embeds SynthID watermarks into every AI-generated image while claiming these markers are "critical" for content authenticity.
But $19.99/month Ultra subscribers can remove the visible watermark entirely, turning transparency into a monetizable product tier. The contradiction runs deeper than pricing.
SynthID verification only detects Google's own images, missing the $347 million in quarterly deepfake losses documented by Resemble.ai from other sources like Midjourney and DALL-E.
The company promises future C2PA standard support, but as of 2025 adoption remains minimal while security researcher Dr. Neal Krawetz demonstrated multiple bypass methods, calling it "snake oil."
Meanwhile, API pricing at $0.24 per 4K image, roughly 5x Midjourney's per-image cost, suggests unit economics requiring subsidies from Search revenue.
Alphabet gained $140 billion in market cap on the announcement as investors rewarded capability demonstrations over sustainable business models. The original Nano Banana generated 13 million users in four days, creating a conversion funnel where free viral growth pushes power users toward paid tiers through quota constraints.
Real authentication problems, from the $25 million Hong Kong CFO fraud to health scam deepfakes, remain unsolved by tools that only verify content voluntarily watermarked by compliant services.
Why This Matters:
• Google monetizes transparency by making watermark removal a premium feature, undermining authenticity as public good while competitors fragment detection standards through 2026.
• Verification theater addresses narrow Google-only cases while actual deepfake threats causing hundreds of millions in fraud losses exploit the unregulated generator ecosystem indefinitely.
Robert Brown
Consensus is an AI-powered academic search engine that instantly searches through 200+ million peer-reviewed research papers to answer your questions. It uses AI to summarize findings, show scientific consensus on topics, and help you understand complex research without reading dozens of papers.
Tutorial:
URL: https://consensus.app/
Brainstorm 3 distinct visual storytelling concepts for a [brand name] campaign centered on [theme].
For each concept, provide:
Additional context:
Ensure the three concepts offer meaningfully different creative directions, not variations of the same idea.
OpenAI CEO Sam Altman warned employees in an internal memo last month that Google's recent artificial intelligence advancements could create "temporary economic headwinds" for the company. The acknowledgment signals growing competitive pressure in the AI sector as Google appears to be making significant progress that could challenge OpenAI's market position.
OpenAI has announced a strategic partnership with Foxconn (Hon Hai Precision Industry Co.) to design and develop data center server racks that will be manufactured in the United States. Under the agreement, Foxconn will produce essential data center infrastructure including cabling, power systems, and other critical hardware components to support OpenAI's expanding computational needs.
Foxconn Chairman Young Liu announced the company will invest $2-3 billion per year in artificial intelligence development over the next three to five years, marking a significant strategic shift for the world's largest contract electronics manufacturer. The Taiwanese company is also exploring potential investment opportunities with Japan as part of its AI expansion plans, while Liu noted that China's electric vehicle market is expected to face consolidation due to intense competition.
Prediction market platforms Kalshi and Polymarket are rapidly expanding their operations as the Trump administration signals a more permissive regulatory environment for gambling and betting markets. Critics are raising concerns that the accelerated growth of gambling opportunities across the United States could lead to unintended social and economic consequences, even as these companies race to capitalize on the shifting regulatory landscape.
IBM and Cisco Systems announced Thursday their collaborative plan to create networks linking quantum computers across long distances, with a proof-of-concept demonstration targeted for 2030. This ambitious project represents a significant step toward developing a quantum internet infrastructure that could revolutionize secure communications and distributed quantum computing capabilities.
Meta's board of directors, including CEO Mark Zuckerberg, has agreed to pay $190 million to settle investor claims related to their handling of the Cambridge Analytica privacy scandal. The settlement resolves allegations that board members failed in their oversight duties during the data breach crisis that affected millions of Facebook users.
The U.S. Securities and Exchange Commission announced Thursday it has dismissed its 2023 lawsuit against SolarWinds Corporation and Chief Information Security Officer Tim Brown, which accused them of hiding cybersecurity vulnerabilities before a major Russia-linked cyberattack in 2020. The closely watched case had sought to hold the software company and its top security executive accountable for allegedly failing to disclose security weaknesses that preceded one of the most significant supply chain cyberattacks in U.S. history.
GE HealthCare has agreed to acquire Intelerad, a medical imaging software company specializing in cloud-based imaging solutions and digital workflow tools, for $2.3 billion. The acquisition will expand GE HealthCare's presence in the outpatient and ambulatory care markets, complementing its existing focus on hospitals and health systems.
Elon Musk's social media platform X has ended its 2023 lawsuit against prominent law firm Wachtell, Lipton, Rosen & Katz, abandoning efforts to recover most of the $90 million fee the firm received for successfully preventing Musk from backing out of his Twitter purchase. The lawsuit had sought to reclaim the substantial legal fees paid to Wachtell for their role in forcing Musk to complete the $44 billion acquisition of Twitter, which he later rebranded as X.
Taiwan's minister Wu Cheng-wen announced that the United States will not impose "punishing" tariffs on Taiwan following a consensus agreement where Taiwan committed to supporting the US chip industry. The tech official stated that Taipei is willing to share expertise from its world-leading semiconductor industrial model as part of a broader trade deal between the two nations.
Samsung Electronics has officially appointed TM Roh, its mobile division chief, as co-CEO and head of the Device Experience (DX) division. Roh had been serving as acting head of the consumer business since April following the death of co-CEO Han Jong-Hee.
Australia's internet watchdog has officially added Amazon-owned live streaming service Twitch to its social media ban for users under 16 years old. In response to the new regulations, Twitch will deactivate existing accounts for users 16 and under starting January 9, 2025, and will prohibit new underage user registrations beginning December 10, 2024.
The UK government has announced a £100 million initiative to support British AI hardware startups through guaranteed "first customer" payments, designed to reduce early-stage commercial risks for emerging companies in the sector. The plan, unveiled by Liz Kendall, comes as the government estimates the UK's artificial intelligence market to be valued at over £72 billion, highlighting the strategic importance of developing domestic capabilities in AI hardware manufacturing.
Archetype AI, a US-based startup specializing in artificial intelligence systems that interpret sensor data from physical environments, has successfully closed a $35 million Series A funding round with participation from Samsung. The company has simultaneously launched new tools designed to help developers build and deploy AI-powered physical agents that can interact with real-world environments through advanced sensor data analysis.
Numerai, the crowdsourcing hedge fund that uses AI-driven trading ideas from freelance quantitative analysts, has successfully raised $30 million in Series C funding led by university endowments, achieving a valuation of $500 million. The Paul Tudor Jones-backed company, which operates a unique model of sourcing investment strategies from independent finance quants, attracted investment from both university endowments and venture capital firms in this latest funding round.
Numerai: The Hedge Fund That Thinks It's Open Source
Numerai calls itself "the last hedge fund," which feels simultaneously grandiose and strangely plausible. This San Francisco outfit pays thousands of anonymous data scientists in cryptocurrency to build trading models they'll never fully understand, then blends them into one meta-model that manages $550 million.
The Founders
Richard Craib, a South African mathematician who cut his teeth at a $15 billion asset manager, launched Numerai in 2015 after getting frustrated with how finance hoards intellectual property. The company employs a surprisingly small team (exact headcount unclear but far below what $550M AUM typically requires) across San Francisco and a recently opened New York office. Craib's pitch: why should tech feel positive-sum while Wall Street stays zero-sum? 🤔
The Product
Numerai runs a market-neutral equity hedge fund that's basically a Voltron of crowd-sourced quant models. Contributors download encrypted financial data, train algorithms without knowing which stocks they're predicting, then stake NMR tokens on their submissions. Good predictions earn tokens. Bad ones get burned. The meta-model aggregates thousands of these predictions into actual trades. In 2024 they posted a 25.45% net return with a 2.75 Sharpe ratio, which is legitimately impressive.
The Competition
Traditional quant giants like Renaissance and Two Sigma play in the same sandbox but keep everything in-house. Quantopian tried the crowdsourced approach and died in 2020. WorldQuant runs massive competitions but doesn't let the crowd drive real money. Numerai occupies a weird middle ground between prediction markets (Polymarket, Kalshi) and conventional quant funds.
Financing
The November 2025 Series C pulled in $30M at a $500M valuation, up 5x from 2023. Backers include Paul Tudor Jones, Union Square Ventures, and several university endowments. JPMorgan committed up to $500M in capacity. Total raised: over $60M across equity and token sales.
The Future ⭐⭐⭐⭐
The model's clever, performance is strong, and JPMorgan's blessing matters. Risks include model herding, regime breaks, and regulatory scrutiny around token incentives. But if Wall Street's future involves tapping global talent without traditional hiring, Numerai might actually be early to something real.
Good Morning from San Francisco, When Congress rejects your proposal 99 to 1, most people take the hint. Silicon Valley&
Good Morning from San Francisco, Trump wants Congress to block state AI laws through December's defense bill. The
Good Morning from San Francisco, When the CEO warning about investment bubbles is spending $75 billion on the very thing
Good Morning from San Francisco, When Jeff Bezos warns about industrial bubbles, listen to his checkbook, not his speeches. When
Get the 5-minute Silicon Valley AI briefing, every weekday morning — free.
