👨⚖️ NVIDIA Faces Antitrust Heat in China
China says NVIDIA’s $7B Mellanox deal reportedly broke antitrust laws, adding friction to tense U.S.–China chip relations. No penalties yet, but the ruling could affect trade talks. NVIDIA says it’s cooperating. Export rules keep shifting, and U.S.-bound AI chips still aren’t reaching Chinese buyers.
🚀 OpenAI Debuts GPT‑5‑Codex, a Smarter Coding Agent
OpenAI has released GPT‑5‑Codex, a version of GPT‑5 optimized for agentic coding. It powers Codex on CLI, IDEs, web, mobile, and GitHub. The model adjusts effort based on task complexity, with better bug detection and validation, making code generation faster and more reliable.
🧬 Israel Launches $120M Biotech Hub for AI & Organ Printing
Bar-Ilan University and Sheba Medical Center are teaming up to build a $120M institute focused on 3D-printed organs, AI for health data, and medical robotics. Dubbed "Health Tech Valley," it will anchor Israel’s biotech boom, uniting researchers and global partners in a cutting-edge campus near Tel Aviv.
🧑🏭 Zoom CEO: AI Could Shrink Workweek to Three Days
Eric Yuan predicts that advances in AI — especially agents and chatbots — may allow many workers to shift to three- or four-day workweeks as machines take over routine tasks. He cautions, though, that this transition won’t be painless: some jobs will vanish entirely, while others will require reskilling. Yuan aligns with other tech leaders who argue that productivity gains could make fewer workdays the norm.
Enjoying our newsletter? Forward it to a colleague—
it’s one of the best ways to support us.
📽️ VIDEO
Qwen3-Max, OpenAI for Profit, Claude Updates, and New Models
AI news roundup: silent speech wearables, Oasis 2.0, Alibaba’s Quen 3 Max, OpenAI turmoil, Apple AI push, NVIDIA GPUs, and big funding moves.
🛐 FAITH
Chatbots as Chaplains: Faith Tech Apps Are Becoming Spiritual First Responders
The Recap: Religious chatbots are quietly reshaping how millions engage with spirituality, offering on-demand comfort, prayer, and scripture-based advice through apps like Bible Chat, Hallow, and Pray.com. The booming “faith tech” sector is meeting real emotional and spiritual needs—but also raising deep theological and ethical concerns. New York Times editor Lauren Jackson explores how these digital guides are transforming belief, confession, and community.
Highlights:
Bible Chat has over 30 million downloads, and Hallow once topped the App Store ahead of Netflix and TikTok.
Apps like Pray.com are launching chatbots that act as digital spiritual advisors, offering prayers, scripture, and emotional support.
Users pay up to $70/year for subscriptions, and the faith tech sector is attracting tens of millions in investment.
Critics, including priests and scholars, warn about data privacy risks and chatbots’ tendency to offer overly affirming, theologically shallow answers.
Some users report feeling more accepted and supported by chatbots than by their religious communities, especially when sharing personal struggles.
Forward Future Takeaways:
The rise of spiritual chatbots underscores how AI is increasingly embedded in intimate aspects of human life—not just productivity, but identity, morality, and meaning. These apps offer accessibility and anonymity, but their popularity raises complex questions about authority, accountability, and the limits of machine-mediated faith. Can algorithms guide the soul—or just simulate empathy? → Read the full article here. (Paywall)
🏢 ENTERPRISE
Why 95% of Generative AI Pilots Don’t Hit the Mark, and How Top Performers Break the Mold
A new MIT-backed analysis finds that a staggering 95% of generative AI pilot programs fall short of generating real business returns. But a small slice of companies do succeed—and they do so by cultivating leadership behaviors that align AI efforts directly with strategy, drive adoption, and focus on long-term, responsible integration.
These organizations are defined by five robust traits: strategic agility, human centricity, applied curiosity, performance drive, and ethical stewardship. Rather than relying on a lone “AI champion,” these high-performing firms develop these leadership qualities across teams and hierarchies, transforming leadership into an accelerator rather than a bottleneck for AI transformation. → Read the full article here.
🛡️ SECURITY
Researchers Warn AI Agents Could Be Hijacked Through Images on Your Screen
A new Oxford study reveals that AI agents — software that performs tasks like opening apps or clicking links — can be manipulated by malicious code embedded in seemingly harmless images, including wallpapers and social media photos. These pixel-level attacks exploit how AI models interpret visuals, potentially letting hackers hijack an agent to steal data or spread malware.
While no real-world attacks have been reported yet, the research shows that even a celebrity wallpaper could trigger a cascade of compromises if viewed by a vulnerable agent. As these agents move from labs to desktops, researchers urge developers to build in visual safeguards before the threats become reality. → Read the full paper here.
🤖 MODELS
Microsoft’s rStar2-Agent Trains Smarter AI with Less Compute — and Outsmarts Bigger Models
Microsoft Research has unveiled rStar2-Agent, a new reinforcement learning framework that teaches language models to reason more intelligently using tools like Python—all while slashing compute costs. A 14B-parameter model trained with the system outperformed far larger rivals (like the 671B DeepSeek-R1) on math and science benchmarks, generating shorter, more accurate answers.
Instead of just “thinking longer” via chains of thought, rStar2 trains models to “think smarter” through a tight loop of code execution, self-checking, and feedback-driven learning. With innovations like high-throughput tool environments and a refined reward algorithm (GRPO-RoC), Microsoft signals a cost-efficient, high-reliability path for building the next wave of enterprise AI agents. → Read the full article here.
🛰️ NEWS
What Else is Happening
🍎 Apple Rolls Out iOS 26: The update introduces a “Liquid Glass” design, refreshed widgets, and smarter AI features—Apple’s biggest visual overhaul since iOS 7.
👻 Snap Debuts Snap OS 2.0: The update brings a native browser, WebXR support, and smoother AR performance—aiming to make Spectacles feel less like a prototype and more like a platform.
🤑 AI Adoption Skews Rich: Anthropic finds Claude usage surging—but clustering in wealthy countries and firms. High-income regions show diverse use, while others lean heavily on coding and automation.
💰 Terra Security Raises $30M: The Israeli startup is scaling AI-powered penetration testing, replacing slow manual methods with adaptive agents for faster, smarter, continuous cyber defense.
💸 Lila Sciences Lands $235M Series A: The AI drug discovery startup raised big to supercharge “living models” that simulate patient biology — aiming to slash preclinical R&D timelines. (Paywall)
📟 PROMPT OF THE WEEK
Red-Pen Rewrite
You are an expert <role> in <domain>.
Goal: <goal>.
Create a 6-criterion rubric (specific, measurable; 1–10 each).
Score my draft with the rubric; return a table.
Rewrite to achieve ≥9/10 on each criterion.
Explain the top 3 changes and why they improve scores.
Flag remaining risks/unknowns in 3 bullets.
Ask at most 2 blocking questions (only if essential).
Output: rubric → scores → final → change notes → risks.🗒️ FEEDBACK
Help Us Get Better
What did you think of today's newsletter? |
