Silicon Valley promised AI would democratize creativity. New research tracking 442 participants found the opposite: people who were more creative without AI produced better work with it. The gap didn't close. It may have widened.
Enterprises are spending billions on AI pilots, but MIT's research shows most deliver no return. It's not the technology failing. The gap between impressive demos and working systems comes down to data quality, technical debt, and organizational readiness.
Good Morning from San Francisco, Anthropic crowned its new model the coding champion. The White House invoked the Manhattan Project
Good Morning from San Francisco,
Anthropic crowned its new model the coding champion. The White House invoked the Manhattan Project for AI science. Both announcements share something: impressive framing that dissolves under closer inspection.
Opus 4.5 leads benchmarks by four points while security tests reveal a 22-point gap between controlled scenarios and real-world agent behavior. The "beat all humans" claim? Only with multiple attempts and cherry-picked results. Genesis Mission promises national lab coordination at unprecedented scale, funded by whatever Congress already allocated. The phrase "subject to available appropriations" appears four times. That's not fine print. That's the whole story.
Two announcements. Same playbook. Declare victory, hedge quietly, let the press release do the heavy lifting.
Stay curious,
Marcus Schuler
Anthropic's third major model in eight weeks arrives with a 67% price cut and benchmark leadership that obscures messier realities. Opus 4.5 scores 80.9% on SWE-bench Verified, edging past Google's Gemini 3 Pro at 76.2% and OpenAI's GPT-5.1-Codex-Max at 77.9%. The margins are narrow at the top.
Pricing drops to $5/$25 per million tokens from $15/$75, shifting Opus from premium positioning toward production viability. But security evaluations reveal troubling variance: 100% refusal rate on one coding test, only 78% when Claude Code faced malware and DDoS requests. That 22-point gap matters for anyone deploying agents with tool access.
The "beat all humans" engineering exam claim relied on parallel test-time compute, running multiple attempts and selecting the best. Without that technique, Opus only matched the top human candidate. Developer Simon Willison tested Opus 4.5 for two days, then switched to Sonnet 4.5 mid-project and kept working at the same pace. Benchmarks measure something. Whether that something translates to daily work remains uncertain.
Why This Matters:
Marcus Schuler
Marcus Schuler
Prompt:
Low-angle shooting，Icelandic scenes, black sand beaches--photographer Tim Walker's attention to detail is astonishing. Photography techniques include: key lighting, panoramic shooting, fashion photography, and commercial photography.
Trump signed an executive order Monday launching the Genesis Mission, a DOE-led initiative to connect 17 national laboratories into an integrated AI platform for scientific research. The phrase "subject to available appropriations" appears four times. That tells you everything.
Officials invoked Apollo and the Manhattan Project in press briefings, but those programs got actual budgets. NASA's funding climbed from $500 million to $5.2 billion in six years. Genesis Mission gets existing supercomputers, existing datasets, and a 270-day deadline to demonstrate results with whatever Congress already allocated.
The named private partners, Nvidia, AMD, Dell, and HPE, had already signed national lab deals before this order existed. Genesis claims credit for work underway. Meanwhile, Energy Secretary Wright promised the initiative would lower electricity prices, even as DOE projects data centers consuming up to 12% of US power by 2028. The administration is simultaneously gutting university research funding while arguing American science needs emergency intervention. A leaked draft included state AI preemption language. It vanished from the signed version, separating the regulatory fight from the science announcement.
Why This Matters
Robert Brown
Guru is an AI-powered knowledge management platform that centralizes your company's information and delivers instant, verified answers where your team works. It connects to your existing tools like Slack, Google Drive, and Confluence, then uses Knowledge Agents to surface trusted information without switching apps. Built-in verification workflows ensure your knowledge stays accurate and up-to-date.
Tutorial:
Act as a McKinsey senior partner developing a 3-year strategic growth plan. The plan should chart a path to 200% year-over-year revenue growth.
Context to incorporate:
– Company: [describe business model, current stage, and market position]
– Current ARR/revenue baseline: [amount]
– Industry: [sector and key dynamics]
– Core constraints: [budget, headcount, geographic focus]
Deliverable structure:
1. Executive summary with key strategic bets
2. Market opportunity sizing and competitive positioning
3. Three strategic pillars with initiatives mapped to each year
4. Revenue bridge showing how growth compounds (Year 1 → 2 → 3)
5. Investment requirements and expected ROI by initiative
6. Key risks and mitigation strategies
7. Critical milestones and decision gates
Format: Use the pyramid principle. Lead with recommendations, then supporting logic. Include one summary exhibit per section.
A new feature on X that exposes the locations of political accounts has demonstrated that trust and safety work was always about combating coordinated inauthentic behavior rather than censoring viewpoints, according to Techdirt's Mike Masnick. The development challenges narratives pushed by critics like Matt Taibbi who characterized content moderation as part of a "censorship industrial complex," showing instead that such efforts targeted bad actors engaging in deceptive coordinated campaigns rather than suppressing legitimate political speech.
Multiple countries including South Korea, the European Union, the United Kingdom, Saudi Arabia, and the United Arab Emirates are pursuing the development of "sovereign AI" capabilities to reduce their dependence on technology superpowers. According to Gartner projections, global AI spending is expected to reach $1.5 trillion in 2025, representing a 50% increase from 2024, as nations with established domestic tech sectors invest heavily in building independent artificial intelligence infrastructure.
New York's proposed RAISE Act would require artificial intelligence companies to publish their safety protocols and disclose serious incidents, positioning the state as a key player in the national debate over AI regulation. The legislation's co-sponsor has become the target of a pro-AI super PAC with ties to the Trump administration and the tech industry, highlighting the intense political battle surrounding efforts to impose oversight on AI development.
According to an internal memo obtained by Reuters, Amazon is directing its engineers to use Kiro, the company's in-house AI coding assistant, instead of third-party alternatives such as Cursor. The move is designed to gather employee feedback that will help Amazon improve and refine its proprietary AI coding tool as competition intensifies in the AI-assisted software development space.
Google is making strides in narrowing the competitive gap with OpenAI by introducing innovative features to its Gemini AI assistant. The company's new "dynamic view" option allows users to convert standard text responses into interactive, visual outputs, representing a significant product enhancement that tech analyst M.G. Siegler highlights as evidence of Google catching up in what he describes as "product delight" — his metric for ranking AI services.
A Google job listing uncovered by Android Authority reveals the company is developing a new operating system called "Aluminium OS," an Android-based platform described as "built with AI at the core" that appears designed to replace ChromeOS on personal computers. The discovery suggests Google is moving toward unifying its operating system strategy by bringing Android to the PC market while emphasizing artificial intelligence as a foundational feature of the new platform.
Swedish buy-now-pay-later giant Klarna announced the launch of KlarnaUSD, its first U.S. dollar-backed stablecoin, built on the Tempo blockchain developed by Stripe and Paradigm. The move positions Klarna to compete in the digital payments space by targeting reduced costs for international transactions, marking a significant expansion of the fintech company's services into cryptocurrency infrastructure.
China is rapidly integrating artificial intelligence and robotics across its manufacturing sector and ports as a strategic response to the Trump administration's efforts to bring global manufacturing back to the United States. Major Chinese companies are leading the development of AI-powered "dark factories"—fully automated facilities that can operate without human workers—enabling faster production and export of goods to maintain the country's dominant position in global manufacturing.
Alibaba Group reported second-quarter revenue of approximately $35 billion, up 5% year-over-year and exceeding analyst estimates of $34.5 billion, driven by a 34% surge in cloud business growth and a 16% increase in Chinese e-commerce revenue. Net income declined to approximately $3 billion as the company increased spending on cloud infrastructure to meet surging demand for AI computing services in China.
Taiwan Semiconductor Manufacturing Co. (TSMC) has filed a lawsuit against former Vice President Lo Wen-jen, who departed the company to join competitor Intel Corp., alleging a high likelihood that he disclosed trade secrets to his new employer. The legal action comes after Lo spent more than 20 years at TSMC, during which time he would have had access to sensitive proprietary information at the world's largest contract chipmaker.
Poland's antitrust authority UOKiK has opened an investigation into Apple to determine whether the company's App Tracking Transparency (ATT) framework restricts competition in the mobile advertising market by limiting third-party data collection while potentially favoring Apple's own advertising services. The probe focuses on whether Apple's privacy policy, which requires apps to obtain user permission before tracking their activity across other apps and websites, creates an uneven playing field that advantages Apple's ad business over competitors.
A new analysis from tech journalist Casey Newton at Platformer examines how technology company executives increasingly employ "delay, deny, and deflect" strategies when confronted with questions about user safety on their platforms. The piece explores the broader implications for tech journalism in an era where public shaming appears to have diminished effectiveness as a tool for holding corporate leaders accountable, raising questions about what mechanisms remain for ensuring platform responsibility.
Character.AI has announced it will cut off access to ongoing chat conversations for users under 18, citing mental health concerns, following the implementation of a two-hour daily usage limit on October 29. The restrictions have reportedly caused significant distress among teenage users, with some expressing deep emotional attachment to their AI conversations.
Character.ai lets you chat with AI personalities, from Socrates to anime girlfriends to custom creations. Think Roblox meets ChatGPT, but for people who'd rather flirt with fictional beings than write emails.
Founders
Noam Shazeer co-wrote the transformer paper that powers basically all modern AI. Daniel De Freitas built Google's chatbot. Google wouldn't ship it, so they left and founded Character Technologies in late 2021. Menlo Park HQ, around 165 employees. Plot twist: Google paid $2.7B in 2024 to license the tech and hire both founders back. 🔄
Product
Pick a character. Talk to it. That's it. Users create bots with custom backstories and personalities, then publish them for others. Features include group chats with multiple AI characters, word games, and multimedia experiments. The company now runs on open-source models like Llama and Qwen. Users average over an hour daily chatting with imaginary friends.
Competition
General chatbots (ChatGPT, Claude, Gemini) eat attention. Companion apps (Replika, Chai, CrushOn) chase the same lonely hearts. Meta embeds AI characters directly into Instagram and WhatsApp. Character.ai differentiates on depth, not breadth. It's the specialized roleplay destination, not the do-everything assistant.
Financing
$43M seed from a16z and angels (2021). $150M Series A at $1B valuation (2023). Total raised: ~$193M. The Google licensing deal implies $2.7B+ effective value, though no new priced round confirms it. DOJ now probing whether that deal dodged merger rules.
Future ⭐⭐⭐
Rough waters ahead. Lawsuits blame the platform for teen suicides. Disney sent a cease-and-desist. Under-18s are now banned from chat entirely. New CEO Karandeep Anand pivots hard toward "entertainment," ditching AGI talk for Netflix comparisons. Revenue hitting $30M+ annually with $50M targeted for 2025. Survival depends on proving fictional characters can grow up faster than their teenage fanbase.
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, 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
Get the 5-minute Silicon Valley AI briefing, every weekday morning — free.
