AI found an Ethereum bug that could take validators offline, but humans had to prove it
AI agents discovered a remotely triggerable crash vulnerability in Ethereum validator software during a coordinated security test by the Ethereum Foundation. The AI generated many confident bug reports, though many were false positives rather than actual vulnerabilities.
Second pregnancy changes the brain in surprising new ways
Each pregnancy produces distinct patterns of brain rewiring rather than repeating the same changes. A second pregnancy alters the brain differently than a first pregnancy, potentially helping researchers develop better treatments for peripartum depression.
The EU says autoplay and infinite scroll are illegal. Meta has until it responds to disagree.
The European Commission issued preliminary findings accusing Meta of using addictive features like autoplay and infinite scroll on Facebook and Instagram. Meta has a formal opportunity to respond before the EU reaches a final decision that could impose fines up to 6% of its global annual revenue.
OpenAI has folded safety into research again. Its head of safety is leaving.
OpenAI merged its safety and research teams under unified leadership, causing safety systems head Johannes Heidecke to depart. The restructuring places safety teams under Mia Glaese, whose title was expanded to cover both domains.
UK’s First Hospital for Houseplants – Where Plants Are Put on Drips and Quarantined for Bugs
A houseplant hospital opened in Edinburgh, Scotland, where plants receive treatment including drip irrigation and quarantine for pest infestations. Rosanna Costello founded the Hilda Houseplant Hospital to help customers maintain their indoor plants' health.
Robinhood Plans to Let AI Agents Trade Crypto for US Customers
Robinhood will allow eligible US customers to connect AI agents to dedicated accounts for autonomous cryptocurrency trading. The company expanded its Agentic Trading product, which originally launched for equities in May, to include crypto assets.
Ethics Deadlock Threatens Senate Crypto Bill Despite July Vote Target
A merged Senate crypto bill aims for a July 20 floor vote but faces a Democratic ethics provision dispute. The central disagreement concerns whether officials should be barred from holding cryptocurrency ties.
Meta killed its Muse Image AI feature three days after launch. Hollywood had had enough.
Meta discontinued its Muse Image AI feature after three days due to privacy concerns with the tool. The model, developed by Meta Superintelligence Labs under chief AI officer Alexandr Wang, was removed from Instagram and the Meta AI app shortly after its Tuesday launch.
Surge pricing hurts gig workers and consumers. Here’s how to fix it
Surge pricing algorithms adjust ride and delivery costs based on demand fluctuations, but research shows drivers receive minimal benefit while consumers face volatile prices. Massachusetts implemented a $34.48 minimum hourly wage for rideshare drivers following a 2024 settlement with platforms.
Why you should be skeptical about financial advice from chatbots
Nineteen percent of surveyed adults reported losing over one hundred dollars following financial advice from AI chatbots. Chatbots provide confident recommendations without acknowledging uncertainty or considering individual circumstances like a spouse's health status that could significantly alter optimal financial decisions.
Apple sues OpenAI for stealing hardware designs, alleging employees brought prototypes to “show and tell” interviews
Apple sued OpenAI, alleging the company used current and former Apple employees to steal hardware designs for AI devices. The lawsuit names OpenAI's chief hardware officer Tang Tan and former Apple engineer Chang Liu as defendants in California federal court.
The Weekender: Honey, I Shrunk the Unicorn
Startup organizational structures have shifted from traditional hierarchies with multiple departments to lean teams supplemented by external tools and AI services. Modern startups now operate with core founders supported by software platforms like Claude, Stripe, and Midjourney rather than full internal teams.
Ant Group’s Robbyant Unveils LingBot-VA 2.0: A Causal Video-Action Model Built Natively for Physical AI
LingBot-VA 2.0 is a video-action model built natively for physical AI that predicts future states before execution and operates at 225 Hz asynchronous control. The model uses Foresight Reasoning to anticipate outcomes and re-grounds on real observations rather than being fine-tuned from a video generator.
Physicists finally build a quantum material predicted more than a decade ago
Researchers created a two-dimensional quantum material with unusual conducting edge states that were predicted over a decade ago. The material's conducting states can be controlled through strain, potentially enabling room-temperature quantum electronics applications.
Tiny bubbles could revolutionize inkjet printing
Researchers demonstrated that varying the quantity of ultra-fine bubbles within droplets can reshape printed patterns during inkjet printing without leaving chemical residues. By adjusting bubble numbers in each droplet, the technology enables cleaner printing for next-generation electronics manufacturing.
Clinical Sinner outclasses Djokovic as he books Wimbledon final place
Jannik Sinner defeated Novak Djokovic 6-4 6-4 6-4 in a Wimbledon semifinal on Friday, advancing to Sunday's final. The defending champion comprehensively beat the 39-year-old in straight sets, ending Djokovic's attempt to win a record-equalling eighth Wimbledon title.
US cybersecurity agency CISA had to build its incident playbook during the incident, agency reveals
A CISA contractor employee uploaded exposed passwords to a publicly accessible GitHub repository, which a GitGuardian security researcher discovered and reported in May. The agency had to develop its incident response playbook while actively managing the breach situation itself.
Geoscientists reveal how Earth's forces are shaping the 'Roof of the World'
Earth's deep interior processes are shaping the Tibetan Plateau's unique topography at its summit. An international team of Chinese and UK geoscientists from the University of Glasgow discovered this through studying formation evidence of this high mountainous area.
Greg Brockman Takes Over OpenAI Product Business After Fidji Simo’s Departure
Greg Brockman, OpenAI's president and co-founder, will oversee the ChatGPT product business after Fidji Simo departed in July. Brockman had already managed product responsibilities since April when Simo took medical leave.
To ancient astronomers, Theta Eridani was brighter for 1,000 years—now we know why
Theta Eridani appeared as one of the 13 brightest stars to ancient astronomers like Ptolemy and al-Sufi, but is significantly dimmer today. Researchers determined the star must have been much more luminous between the second century A.D. and A.D. 964 to match historical brightness records.
US CBDC Ban Set to Become Law at Midnight
A four-year ban on U.S. central bank digital currency takes effect Saturday if the housing bill becomes law without presidential veto. President Trump has neither signed nor vetoed the bill that Congress approved and sent to him.
New sensors capture warning signs before fish deaths in Lake Victoria
Sensors installed in Lake Victoria detected low-oxygen warning signs several hours before fish deaths occurred, providing crucial advance notice of dangerous environmental conditions. The research from Kings College London researchers highlights the capability of monitoring equipment to identify these hazardous events before widespread mortality happens.
Apple Sues OpenAI for Trade Secret Theft
Apple sued OpenAI for trade secret theft, claiming the company encouraged Apple employees to share information about upcoming products, components, and drawings. OpenAI's hardware chief was accused of coordinating the campaign to obtain confidential Apple materials.
Finovate Global Israel: Autonomous AI and Cross-Border Payments
Tangos AI, an Israeli fintech specializing in autonomous AI for financial crime and fraud investigations, raised $20 million in seed funding. The funding round was led by Red.
Nubank Unit Secures Authorization to Become Mexican Bank
Nu Mexico received authorization from Mexico's financial regulator to operate as a bank and has 30 calendar days to complete its transformation. The company previously operated as a digital bank before obtaining approval from the National Banking and Securities Commission.
57% of enterprises have watched AI agents be confidently wrong. The fix is an agentic context layer, but who has one?
Fifty-seven percent of enterprises reported AI agents providing confidently incorrect answers due to missing or inconsistent business context in the past six months. Thirty-eight percent of enterprises rely on document retrieval as their primary method for giving agents business context, while seventy-five percent lack a dedicated agentic context layer to govern this information.
OpenAI introduces ChatGPT Work, a cloud-based AI agent that manages tasks across email, Slack and calendars
OpenAI launched ChatGPT Work, an AI agent powered by GPT-5.6 that executes complex multi-step tasks across email, Slack, calendars, and code repositories. The agent can gather context from connected apps to produce finished documents, spreadsheets, presentations, reports, and websites while working independently on projects for hours.
Kaiser nurses say AI is changing their jobs—for the worse
Kaiser Permanente implemented AI monitoring that evaluates nurses based on call duration, with calls exceeding 15 minutes triggering reprimands or performance score impacts. Nurses protested that this system forces them to follow scripts and limit advice, potentially compromising patient care quality.
Delta AI Assistant Boosts Customer Satisfaction Scores 25 Points During Travel Disruptions
Customer satisfaction scores increased 25 points during travel disruptions through simplified rebooking processes and expanded self-service digital tools. Delta Air Lines implemented these enhancements, which the chief operating officer discussed during the company's June quarter earnings call.
LinkedIn is the ‘most AI-saturated platform,’ new study suggests
More than 40 percent of LinkedIn posts longer than 250 words were flagged as fully AI-generated, according to an AI detection study that scanned nearly 1 million posts across social platforms over two months. LinkedIn posts accounted for nearly two-thirds of all AI-flagged content despite representing only one-third of the scanned posts.