Generative AI — Posts

56 posts on "Generative AI" by Shlomo Strauss — bilingual writing on technology, science, and business.

• How AI threatens reading yet enables custom printed books — As visual media and AI reduce our reliance on text, generative tools offer a paradoxical solution by creating highly personalized physical books.
• The dopamine trap of building useless AI projects — The ease of generating code with AI creates a dopamine loop of unused projects, highlighting the need to focus on fundamentals and real value.
• The structural flaws driving OpenAI toward collapse — Without its own infrastructure, OpenAI's growing user base deepens its debt, risking a collapse that could trigger a broader tech recession.
• Corning's evolution from Edison's lightbulb to AI optical fibers — Corning has sustained 175 years of relevance by adapting its glass technology for everything from Edison's lightbulbs to modern AI data centers.
• An autonomous AI system turns bug reports into deployed code — A newly developed AI reporting tool analyzes user feedback to automatically provide tech support, write code, and deploy software updates.
• An AI-native game engine is taking aim at Epic and Unity's duopoly — Immense, a new game engine built by a former Epic director, uses parallel AI agents to handle physics, lighting, and audio — and could disrupt the…
• Google Reimagines the Laptop with Android-Powered GoogleBooks — Google's upcoming GoogleBook laptops will run a redesigned Android OS with native app support and deep Gemini AI integration to challenge Windows.
• An Island Experiment in AI Governance and Digital Twins — British startup Sensay is launching a social experiment on a tropical island, using AI digital twins of historical figures to govern a new state.
• How Cerebras Built a Giant AI Chip and Why It Is Going Public — While Cerebras boasts revolutionary wafer-scale AI chips, its upcoming multi-billion dollar IPO faces risks from extreme customer concentration.
• Nvidia's $9 Million Server Threatens Traditional Hardware Makers — Nvidia's upcoming NVL144 server highlights the company's push to control the entire AI supply chain, threatening traditional hardware manufacturers.
• Google's TurboQuant Algorithm Cuts AI Memory Needs by 8x — Google's open-source TurboQuant algorithm uses polar coordinates to compress AI memory requirements by eightfold without losing accuracy.
• Agentic browsers and open-source platforms make a surprisingly — Pairing mature open-source platforms with an agentic AI browser bridges the gap between an empty self-hosted stack and a fully configured, working system…
• Why Musk's controversial xAI moves may follow a familiar playbook — The Tesla-to-xAI money flows look like investor betrayal, but Musk's Solar City precedent suggests a ruthless rebuild strategy rather than a bailout.
• Fiber-optic cables and AI are finding hidden water leaks underground — A British startup repurposed existing fiber-optic infrastructure and AI models to detect water pipe leaks, saving over 700 million liters of water…
• Anthropic vs. the Pentagon: a $200M contract and a question of who — After Anthropic refused to lift Claude's safety restrictions for the Pentagon, losing a $200M contract, the clash raises hard questions about corporate…
• Why old HDD drives are suddenly back in high demand — EU AI data-retention rules, rising SSD costs, and new laser and microwave writing technologies are driving an unexpected HDD revival with capacities up to
• Intel's ZAM memory aims to pack 512 GB where HBM tops out at 64 — Intel and SoftBank's ZAM technology uses diagonal drilling between memory layers to promise 512 GB per package — eight times the capacity of today's best…
• Musk's AI empire: custom chips, space servers, and a closed-loop — From Dojo 3 supercomputers in orbit to xAI merging with SpaceX, the piece maps how Musk is assembling a self-contained AI pipeline that no competitor…
• Google's AI game demo wiped billions off gaming stocks — but is the — Google's Project Genie demo triggered a sharp sell-off in major gaming stocks, but the real AI threat to the industry likely comes from code-generation…
• How Nvidia quietly sought pirated data from Anna's Archive to train — An internal Nvidia email revealed the company pursued access to Anna's Archive's illegal dataset for AI training, exposing the ethical compromises driving
• Why Nvidia's $20 billion Groq acquisition is a bet on the inference — As the AI industry shifts from model training to cost-efficient inference, Nvidia's Groq acquisition signals a strategic pivot to compete with Google's…
• How data center water use compares to the beef industry — A breakdown of xAI's Colossus 2 data center water consumption against hamburger production figures challenges the assumption that AI infrastructure is…
• Why Nvidia skipped new GPU launches and what DLSS 4.5 signals instead — VRAM shortages driven by AI server demand blocked new GPU launches, so Nvidia responded with a Transformer-based DLSS 4.5 upgrade that generates graphics…
• Biological mini-brains as computers: inside FinalSpark's experiment — Swiss startup FinalSpark grows human brain cells on electrodes and trains them with dopamine to perform computing tasks at roughly one-millionth the…
• Using AI to generate a sequence diagram from real production code — By connecting a GitHub repository to Google's Firebase Studio and Mermaid, the author generated a detailed sequence diagram that maps a web scraper's full
• Why Google may win the AI race — and it has little to do with its — An analysis of the four economic fundamentals — chips, cloud, data, and ecosystem — that will determine which AI companies survive, and why Google alone…
• Superbugs, antibiotic resistance, and the AI hunting new cures — Overuse of antibiotics, especially in China, is accelerating the rise of untreatable superbugs — and AI models trained on billions of molecules may be…
• How AI's appetite for HBM memory is driving up the cost of ordinary — As manufacturers shift production lines to the more profitable HBM chips powering AI accelerators, a RAM shortage for standard computers is expected to…
• Databricks: the $100 billion startup most people have never heard of — Valued at up to $130 billion, Databricks dominates enterprise AI and big data infrastructure while remaining almost invisible to the general public.
• Jevons Paradox and Baumol's Effect: what AI will do to the economy — As AI cuts the cost per token, Jevons Paradox predicts demand will surge far beyond current levels, while Baumol's Effect explains why your plumber's…
• PROMPTSTEAL: the malware that uses AI to build itself after — Google researchers exposed a Russian-linked malware that carries no malicious code upfront, instead querying an AI model post-installation to generate…
• Credo Technology: the cable maker quietly powering AI data centers — Credo's active electrical cables — each carrying onboard chips that clean and amplify signals — have made it a near-$25 billion company fueling the…
• Building an automated Telegram news channel in under two hours — Using RSS feeds, a self-hosted N8N server, and a lightweight AI model, the author automated the translation and publishing of tech and finance headlines…
• 10 principles behind a consistent LinkedIn content presence — A technology consultant ranked among LinkedIn's top content creators shares the ten practical principles — consistency, no AI, no selling — that shaped…
• Fake plant seeds generated by AI are selling online for real money — Scammers are using AI-generated plant images to sell seeds for species that never existed, earning thousands of dollars and eroding trust in real…
• Robotic sculpture and the question of what human artistry still means — As Monumental Labs deploys autonomous robotic arms to carve stone, the post asks whether machine-made sculpture can ever carry the same weight as a…
• The 1958 physics experiment that launched a $220 billion industry — A 1958 analog tennis game built on a repurposed oscilloscope is recognized as the first real-time video game, and the unlikely ancestor of today's gaming,
• What a pasta machine reveals about how neural networks see images — A pasta machine splitting one image into smaller copies turns out to be a surprisingly clear model for how convolutional filters and AI image generation…
• The OpenAI model built to add a decade to human lifespan — OpenAI partnered with Retro Biosciences to build a protein-simulation model that could improve cellular reprogramming success rates 50-fold, advancing a…
• Jensen Huang sees billions of robots filling the labor gap within 20 — In a Bloomberg interview, Nvidia's CEO outlined a future shaped by robot labor, physics-based AI training environments, and fully AI-managed data centers.
• Automating SEO for thousands of products with AI and Make — A Make automation pipeline using ChatGPT or Claude rewrites product titles and descriptions at scale, updating thousands of store listings in one click…
• Google's Genie 2 generates playable game worlds from a single image — Google's Genie 2 turns a single image into a navigable game environment, and may soon help studios cut the years and millions required to ship a AAA title.
• Peter Thiel: the tech billionaire behind PayPal, Palantir, and much — A profile of Peter Thiel's career spanning PayPal, Palantir, the first outside investment in Facebook, and a fellowship that helped launch Ethereum and…
• Inside Elon Musk's GPU cluster: how large language models are trained — A step-by-step breakdown of how language models learn — using Musk's massive GPU cluster as a starting point to explain forward passes, loss calculation,…
• Google's Add Me feature puts the photographer back in the group photo — Google's new Add Me feature for Pixel devices uses AI to merge two separate shots into one group photo that includes everyone, even the photographer.
• Google's AI tool lets anyone design a custom font from scratch — A new AI-powered tool from Google Research lets designers and developers generate entirely new typefaces, opening up broad possibilities for web…
• The bookmarks bar is old and unglamorous — and it's your best — Organizing your bookmarks bar into folders and subfolders eliminates the daily time lost to repeatedly searching for the same sites and files.
• Apple's new iPad Pro is impressive — and almost nobody needs to — Five takeaways from Apple's iPad event: why a product that's too good becomes a business problem, what the new OLED display actually delivers, and where…
• Apple's on-device AI strategy may finally give Apple Silicon a reason — By betting on privacy-focused, on-device AI with its new openELM model, Apple may have found the killer use case that finally justifies its powerful…
• Why Apple is paying Shutterstock $50 million for AI training data — Apple's deal with Shutterstock is the latest sign that data ownership has become a defining economic advantage in the AI era, reshaping the value of…
• Aging bodies, robotic exoskeletons, and the future of elder care — With social security projected to turn cash-negative by 2036, the author argues that AI-driven robotics and exoskeletons may render traditional nursing…
• Retail Risks and the Hidden AI Value in Reddit's Unusual IPO — Reddit bypassed institutional investors to offer shares directly to users, but the company's true long-term value may lie in its AI training data.
• CPU vs GPU: why core count matters more than speed — A clear breakdown of how CPUs and GPUs differ in architecture, and why massively parallel but simple cores give GPUs the edge in gaming, crypto, and AI.
• Why Musk's lawsuit against OpenAI may have real strategic logic — Beyond the public spectacle, Musk's lawsuit against OpenAI may be aimed at securing historical credit, source code access, or compensation for his…
• Apple's canceled car and the human cost of a decade-long project — Apple shut down its autonomous car project after a decade and $10 billion, prompting a reflection on what it means for 2,000 engineers who watched years…
• Why your AI startup idea is already obsolete — Data concentration among giants like Google and OpenAI makes most AI startups unviable, and the real opportunity lies in adopting the technology rather…