AI&Future Tech

🚀 Substack Adds AI-Detection Tool Powered by Pangram Substack is rolling out an AI text detection feature across its platform, built in partnership with AI detection company Pangram. Users can scan posts, notes, replies, and comments over 100 words via a "Scan for AI text" option in the post menu. The tool estimates how much content may be AI-generated or AI-assisted. It's live on web and iOS, with Android coming soon. Source

🤖 AI Use Raised Homework Scores But Cut Exam Results, Study Finds A study of 26,800 high school students found that AI assistance boosted homework scores by 18% while exam results dropped by 20%, suggesting students are offloading learning rather than reinforcing it. Source

🤖 Google Ships Three Gemini Flash Models, 3.5 Pro Still Delayed Google released Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber this week — all targeting speed and cost efficiency for AI agents. The 3.6 Flash reduces token usage by up to 17% vs. its predecessor. Flash Cyber is a cybersecurity-focused model restricted to governments and select partners. Gemini 3.5 Pro, originally promised for June, remains in testing with no firm release date. Google also confirmed it has begun pre-training for Gemini 4, its next flagship model. Source

💼 Microsoft and Mistral Sign Multibillion-Dollar Europe AI Deal Microsoft and French AI company Mistral have expanded their strategic partnership with a multibillion-dollar agreement to scale AI infrastructure across Europe. Source

💻 CS Enrollment Drops for First Time in ~20 Years Computer science enrollment fell 8.1% year-over-year in fall 2025, per National Student Clearinghouse data — the first decline in roughly two decades, according to Stanford economist Jacob Light. Light's study, covering 50M+ course sections across 1,019 US colleges since 1996, notes the drop coincides with AI's rise but stops short of attributing causation. Possible factors include shifting expectations about CS career value, labor market conditions, and AI's impact on coursework. Source

💻 TSMC Raises Chip Prices, Piling More Pressure on Apple Apple, which recently raised prices on most products citing higher memory costs, now faces additional cost pressure after TSMC increased its chipmaking prices. iPhone price hikes are still expected when new models launch in September. Source

💼 OpenAI's ChatGPT Ad Business Faces Steep Skepticism OpenAI launched ChatGPT ads in February and has reached $100M in annual recurring revenue, but analysts are dismissing its projections of $2.4B in 2026 and $100B by 2030. EMARKETER's Nate Elliott said the assumptions required "are not logical or reasonable," noting OpenAI would need to beat Google in consumer AI and flood ChatGPT with ads to get there. The entire chatbot ad market is forecast to generate under $1B this year; Google made $294B in ads in 2025. Advertisers say the platform lacks campaign creation, measurement, and reporting tools that Google and Meta offer. Ads currently run in only seven countries, and geographic targeting within the US only became available in July. Source

💼 Anthropic Settles AI Copyright Suit for $1.5B A federal judge in San Francisco has approved a $1.5 billion settlement between Anthropic and authors who sued the company for using their books to train Claude — the largest known copyright settlement in U.S. history. The case was the first major AI copyright lawsuit to reach a conclusion in the United States. Source

💼 Sony Sues Udio Over 30,000 Copyrighted Songs Sony Music has filed a new lawsuit against AI music generator Udio in a New York court, alleging copyright infringement across more than 30,000 songs — including tracks by Elvis Presley, Beyoncé, and Harry Styles. Sony says the list covers "only a small portion" of the works Udio allegedly infringed. The suit follows a 2024 case Sony filed alongside Universal Music Group and Warner Records against both Udio and Suno. Source

💻 Google's "Frozen v2" Chip Embeds Gemini Into Silicon Google is developing a server chip codenamed "Frozen v2" that bakes the Gemini AI architecture directly into hardware. Internal sources say it could be 6–10x more efficient than current TPUs, potentially cutting inference costs significantly. Deployment is targeted for 2028. Source

🤖 AI Job Titles Spread Far Beyond the Tech Sector New Indeed data shows "AI-touched" job titles in the US have more than tripled since 2022. In five of six major markets studied — the US, Germany, France, the UK, and the Netherlands — most AI-labeled roles now sit outside the tech sector, spanning sales, HR, legal, healthcare, and trucking. Spain was the sole exception. The trend signals AI adoption becoming a broad workplace shift, with employers seeking workers who use AI tools rather than build them. Source

🤖 Moonshot and Alibaba Challenge US AI Leaders With New Models Beijing-based Moonshot AI unveiled Kimi K3, claiming it outperforms nearly every US model except OpenAI's top system. Alibaba followed with its own competitive release. Both companies say their models match leading US systems at significantly lower cost. Source