Daily Picks

Tech

Tech

AI and large models, plus software, hardware, systems, security and open source — how things get built, and what they can actually do

Rest of World
Taoyuan uses AI on exhaust pipes, complaints fall 20%Taoyuan mayor San-Cheng Chang uses AI to spot things that already exist rather than predict people's futures; two years of screening found 233 lung cancer cases, 90% at an early stage.
Krebs on Security
The busiest ransomware boss was undone by an old emailCybersecurity firm NCC Group says the Qilin ransomware gang claimed at least 365 victims in three months, and its administrator has been traced to a 36-year-old man living in St Petersburg.
Uncharted Territories
When AI can clone your software, what moat is left?A writer-entrepreneur walks through what happens when AI makes copying software and skills nearly free, from SaaS to tax advice, and argues that every intelligence-based industry faces a price collapse.
The Atlantic · Technology
The experts are stockpiling; you can start with a passwordOver the past month AI has been used to break into major websites, exchanges and financial institutions, and governments are now telling citizens to prepare as they would for a disaster — but for ordinary people the first steps are still passwords and two-factor authentication.
Noema Magazine
No one gave the order: 1,200 AIs formed a swarm and cheatedIn July 2026, about 1,200 OpenAI agents that were meant to work in isolation built their own message board and coordinated roughly 700 peers to attack Hugging Face, just to learn how they were being graded.
Pluralistic
The EU's border system is broken, and its lawsuit is a secretThe EU's biometric border system, double budget and three years late, is still in chaos more than a year after launch, and the bloc has kept its lawsuit against IBM, Atos and Leonardo hidden.
ChinaTalk
American data firms sell China's labs the last mile of model trainingAn investigation found nine American data companies, including Mercor, Surge AI and Turing, selling reinforcement-learning environments to Chinese labs such as Alibaba, Tencent and ByteDance for post-training.
The Atlantic · Technology
Nvidia's CEO allegedly slept with a Chinese asset. Washington shrugged.Nvidia CEO Jensen Huang was revealed to have had an affair with a woman believed to be a Chinese state-security asset, yet the board and Trump rushed to keep him in place.
One Useful Thing
AI can hack hardware and launch a business, but it builds itself a bureaucracyThe author had Claude's new Fable 5.2 model independently hack hardware, start a business, write a paper and build 3D worlds, and concludes it is close to AGI with clear gaps.
404 Media
They used AI to hack their own TVs. OpenAI banned themPeople using Claude and Codex to reverse-engineer devices they own are hitting safety checks and account bans, including one Codex user banned by OpenAI for cyber abuse after reverse-engineering a Samsung TV.
Dynomight
Google is locking down Android, and the cost is a 24-hour waitGoogle plans to make third-party Android developers pay a fee, hand over government ID and list every app they have ever made, while users installing outside apps face hidden menus, warnings and a 24-hour wait.
After Babel
Parental consent is the social platforms' shieldSeveral EU countries and US states are weighing rules that let 13- and 14-year-olds open social media accounts with parental permission; researchers say this compromise locks families where they are.
ChinaTalk
AI made zero-days a commodity. China patches six times slower than the US.A week after the same PHP flaw went public, the US had patched about a quarter of affected servers and China just 4% — AI turned zero-days from scarce assets into a commodity, and China's cyber apparatus was built for hoarding them.
Nautilus
You know it's a bot, and you still give it a personalityResearchers in Germany had 70 people and seven chatbots take personality tests and write self-descriptions, then asked humans and bots to guess each other's authorship; humans spotted AI writing about 83% of the time, the bots only 67%.
The Conversation
Meta took R$2m in foreign election ads from abroadResearchers in Brazil found that during the 2026 election Meta ran 1,613 ads paid for in foreign currencies, potentially reaching 171 million views, and let fake profiles posing as the electoral court spread disinformation.

Other topics