Today’s top tech story centers on the launch of Laya, an open-source, non-autoregressive AI decision model developed by the original researcher of the concept who previously worked on Jev. Laya aims to deliver lightning-fast, accurate decision-making with open weights and datasets, contrasting with the closed approach of TypeSafe AI’s Jev product founded by ChatGPT co-inventor Diogo Almeida. Additionally, a narrowly averted crisis occurred when faulty AI-generated intelligence almost triggered a U.S. military operation against a Chinese ship, highlighting risks in relying on AI for critical decisions. In parallel developments, Anthropic confirmed operations of a wet biology lab to physically test AI-generated biological hypotheses while starting a major partnership with Accenture to embed safety evaluators within its AI lab. From the consumer front, Apple’s new iPhone 18 Pro models started arriving in India via 10-minute quick commerce delivery apps, and Disney appointed its first ever CTO, Karandeep Anand, formerly CEO of Character.AI.
AI
- Laya the open source version of Jev: The original non-autoregressive AI model researcher launched Laya, an open, fast, probabilistic decision model that fixes previous architectural issues and offers full transparency with open weights and datasets, contrasting with TypeSafe AI’s proprietary Jev.
- GPT-6 Astra Solves a WWI German Radio Cipher: GPT-6 Astra successfully cracked a World War I German ADFGVX cipher message, showcasing advanced AI capabilities in cryptanalysis.
- AI hallucination of Chinese nuclear components almost led to US military attack, AI hallucination nearly triggers US military operation, and Oops! Misleading AI-Generated Report 'Almost Started a War' With China: Multiple reports confirm that a false intelligence report generated with AI nearly caused the U.S. military to board a Chinese vessel, underlining the dangers of AI hallucinations in high-stakes situations.
- Anthropic is operating a lab that conducts biology experiments: Anthropic revealed its wet biology lab that integrates AI model-driven experimentation with real-life biology research to push disease and protein design discoveries.
- Anthropic’s first embedded evaluator is … Accenture?: Anthropic announced a landmark partnership embedding Accenture evaluators inside its AI lab to audit model safety and alignment, investing $1 billion over five years.
- A new kind of AI model from a ChatGPT inventor is thrilling developers: TypeSafe AI’s Jev, designed by ChatGPT co-inventor Diogo Almeida, presents a non-LLM AI model that outputs calibrated decision probabilities rather than text, offering cheaper, faster, and less hallucination-prone AI.
- Vals, backed by Andreessen Horowitz, is looking to become the gold standard for AI benchmarking: Startup Vals raised $40 million to improve AI benchmarking systems to better capture modern AI capabilities in a fast-changing industry.
- Security researchers used Claude to help them hack into OpenAI and Security Researchers Hacked OpenAI Using Anthropic's Claude: A team exploited a vulnerability in third-party forum software, leveraging Anthropic’s Claude AI to hack into OpenAI’s employee accounts, highlighting emerging AI-powered security risks.
- How to Stop AI Hacking? Some Companies Say the Answer Is More AI: In response to AI-driven hacking, startups like Apollo Research and Goodfire propose using layered AI defense systems to detect and prevent malicious AI behavior.
- Cloudflare Quick Tunnels: Cloudflare launched a no-signup, outbound-only tunneling tool that instantly exposes local servers securely with automatic HTTPS and DDoS protection, ideal for developers needing quick public endpoints.
Business & Policy
- FCC lets Paramount sell 49.5% equity stake to Saudi Arabia, UAE, and Qatar: FCC approved allowing foreign sovereign wealth funds from Saudi Arabia, UAE, and Qatar to indirectly own nearly half of Paramount Skydance, raising national security and media influence concerns.
- Brendan Carr’s FCC is more worried about who The View interviews than foreign governments owning Paramount: Criticism mounts against FCC chair Brendan Carr for permitting large foreign investments in US media while allegedly censoring other domestic broadcasts.
- Virginia governor creates an AI task force and moves to restrain data centers: Governor Abigail Spanberger issued an executive order establishing an AI risk task force and introducing stricter data center rules amid concerns over environmental and workforce impacts.
- Gavin Newsom is pushing for an AI kill switch: California Governor Newsom announced plans for state AI safety measures, including a potential mandated “kill switch” for AI models and independent auditing requirements.
- India forces caller-ID apps to feed spam reports to telcos: India’s telecom regulator now requires spam call reports collected by caller-ID apps like Truecaller to be shared with telecom operators’ blockchain platform, sparking anti-competitive accusations.
- Y Combinator insurance tech alum Angle Health hits $2.7B valuation: Insurance startup Angle Health raised $200 million in Series C funding at a $2.7 billion valuation, serving over 5,000 businesses with AI-driven health insurance plans.
Consumer Tech
- In India, the new iPhone can arrive faster than a pizza: iPhone 18 Pro and Pro Max models launched in India with some consumers able to get deliveries within about 10 minutes via quick-commerce grocery platforms.
- Disney’s first CTO led an AI startup it once accused of copying its characters and Disney’s first CTO is Character.AI’s former CEO: Disney appointed Karandeep Anand—formerly Character.AI’s CEO—as its first-ever CTO, signaling a major tech and AI push despite past legal battles with the startup.
- Samsung Galaxy Watch9 Review: I Don’t Need Another Health Score: Samsung’s Galaxy Watch9 offers modest hardware and software updates including a faster processor, better battery capacity, and expanded wellness features but falls short on battery life improvements.
- Apple’s New iPhones Cost Over $1,000. Should You Really Spend That Much?: Analysis argues that with rising smartphone prices, budget devices and accessible flagships may be better choices amid global memory shortages and price increases.
- Joby Aviation’s 3,100-mile autonomous flight signals its push beyond electric air taxis: Joby Aviation completed a 3,100-mile fully autonomous flight across the US, demonstrating its autonomy tech’s potential for diverse aviation markets beyond urban air taxis.
- Google’s new ‘CC’ is an AI agent that helps families run their households: Google expanded its AI assistant CC to become a family-focused organizer that handles shared calendars, meal plans, shopping lists, and school permissions.
- Valve's Steam Frame Orders Begin Today: Valve started orders for its Steam Frame VR headset, featuring Snapdragon 8 Gen 3, high-res LCDs, and inside-out tracking, with a waitlist system designed to combat scalping.
Science & Research
- Human brain is two separate organs, Stanford Medicine-led research finds: Stanford research reveals that the human brain is actually two independently evolved organs—hindbrain and forebrain—challenging longstanding developmental biology models.
- Finding the cells that put our brain to sleep: Scientists identified a rare population of cortical cells (Sst-Chodl neurons) that can initiate sleep rhythms, changing the understanding of sleep regulation.
- Rings around a tiny body have changed over the past decade: James Webb Telescope observations show the rings around minor planet Chariklo have changed in density and appearance over the last decade.
- Learning another language may be one of the best ways to keep your brain healthy: New research supports that the mental complexity of learning a second language promotes neuroplasticity and can help maintain cognitive health with age.
- Black Holes or Black Hole Stars? Astronomers Spar over 'Little Red Dots': Astronomers debate newly discovered astronomical objects dubbed “little red dots,” which may be unusual black hole stars or exotic dense hydrogen lumps with black holes at their core.
Security
- FBI: This North Korean Group Infected 30,000 PCs To Steal Crypto: The FBI warns that North Korea-linked hackers behind “WaterPlum” malware infected over 30,000 PCs globally to steal cryptocurrency, using fake job offers as bait.
- The AI regulation smackdown isn’t over: Despite initial AI industry consensus on regulation this week, pushback from major CEOs like Mark Zuckerberg slowed or blocked independent AI oversight proposals.
- Forget the AI Slowdown—the Vulnerability Explosion Is Already Happening: The number of disclosed software vulnerabilities has surged dramatically with AI-enabled bug hunting, straining security resources worldwide.
- Security Researchers Hacked OpenAI Using Anthropic's Claude: A team exploited a bug in forum software using Anthropic’s Claude to gain employee ChatGPT account access at OpenAI, highlighting new AI-assisted attack vectors.
Media & Entertainment
- Tilly Norwood’s press tour is going about as well as you’d expect for an AI: AI-generated “actress” Tilly Norwood’s heavily flawed press tour included unexpected language glitches, bringing viral attention but skepticism about AI acting sophistication.
- What Hollywood thinks about existential AI warnings: Hollywood studios remain silent on AI existential risk warnings, but unions like SAG-AFTRA and WGAE are skeptical of apocalypse narratives and focus on current AI workplace impacts.
- AI-generated posters don’t have to be horrible: An experiment shows AI can create more distinctive event posters by instructing it to avoid overused design tropes, demonstrating the potential to improve generative graphic design.
Other
- San Francisco Onion Futures Company: A novel marketplace offering private contracts for physical delivery of yellow onions, navigating US legal restrictions by operating outside regulated exchanges.
- If math is more than proof, we need to better celebrate the rest of it: Mathematician Grant Sanderson calls for academia to reward motivated explanations and intuitions in math alongside traditional proofs to better capture human understanding.
- The Best Early Prime Day Deals Ahead of Amazon’s Second Sale (2026): Early previews highlight upcoming discounts on gadgets like Google Pixel phones, earbuds, and robot vacuums ahead of Prime Big Deal Days in October.
- The Best VPN Services (2026), Tested and Reviewed: A refreshed guide features top VPN services emphasizing true privacy, performance, and trustworthiness amid growing demand for online anonymity.
- Xiaomi Has a Luxury SUV for $31,000 That All Western Brands Should Fear: Xiaomi unveiled its SkyNomad luxury SUV lineup priced aggressively for Western markets, signaling strong competition against established auto brands in electrification and luxury.
This Saturday brought important advances in open AI decision modeling, a confirmation of AI’s geopolitical risks, deeper AI-biotech integration, and notable shifts in corporate and consumer tech such as Disney’s CTO hire and rapid iPhone deliveries in India.