Discussion | Link

More screen time among children and teens is linked to higher risks of heart and metabolic problems, particularly when combined with insufficient sleep. Danish researchers discovered a measurable rise in cardiometabolic risk scores and a metabolic “fingerprint” in frequent screen users. Experts say better sleep and balanced daily routines can help offset these effects and safeguard lifelong health.

by Dan Shipper in Chain of Thought Midjourney/Every illustration. Was this newsletter forwarded to you? Sign up to get it in your inbox. Over the past weeks we’ve looked at the dominance of the old Western worldview and where it reached its limits. We’ve talked about the potential of language models to help usher in a new worldview, and how this new worldview changes how we might approach science, business, and creativity. We’ve talked about the importance of connectedness and context, participation and paradox, and to close the series, I want to elaborate a bit on the core premise, that our tools shape how we see the world. But first, I want to share more of my own personal context. As a teenager, the one place I felt the most myself was in my room. I could go to my room and write, or code, or run experiments, and no one could tell me what to do. It’s from this room that I started my first internet business in high school, called Convenience Software, selling Blackberry and iPhone apps. I coded at night and spent lunch at school in the library answering customer service emails. That’s how I paid for gas and food. My room is also where I fell in love with writing. When I wrote something—diary entries, short stories, little essays—I could pin down the world. I could know what was true about myself. That desire to pin things down, to make them permanent and knowable, continued as I got to college. I studied philosophy, and each week I was enchanted by a famous system of thought, like Plato’s forms or Descartes’ mind-body dualism, and felt sure that it explained everything. Then the next week, I would encounter a thinker who came just after them—Aristotle or Hobbes—and find that they effortlessly tore down the previous system, which had seemed so perfect, and replaced it with their own. But I kept reading anyway, hoping to find a philosophy that would last. During this time, I also started my first real startup, and I loved it. I was also gripped the whole time by the feeling that I didn’t know enough. I read voraciously and sought advice wherever I could get it. I talked to my dad, who ran the family cemetery and funeral home business, almost every day on the phone. I thought of myself as an information processing machine: If I could just know more about the world, more about business, more about technology, then I could make my company succeed. But I constantly dreaded that the things I learned would slip through my grasp. If I could only record them, I felt, then I’d have them when I needed them. So I began taking copious notes. In a little black notebook I kept in the back of my jeans, I recorded everything from meetings to what books I was reading to my daily spending. I grew my company for a few years and sold it (I flew straight from my college graduation ceremony to Boston to finish negotiating the deal), then spent years figuring out what I wanted to do next. I traveled the world, worked at an incubator, and invested in startups. All the while I had even more time to refine my note-taking system. But the perfect system remained elusive. Each new tool or method I tried seemed promising at first, but each one broke as I tried to use it for new things. I bounced between Evernote, Notion, and countless other apps, creating increasingly complex workflows. It felt like I was jumping from one fad diet to another, just as I had jumped from one system of philosophy to another as an undergraduate. I had a hunch that there was something backwards about these organizational systems. The right system depended very much on what you wanted to use the information for, but note-taking is implicitly for information whose use is open-ended—that’s why notebooks come with blank pages instead of preprinted agendas. I became convinced that if that perfect tool wasn’t already out there, I could build it. I envisioned a system that would adapt to how people actually think and work, rather than forcing them into rigid structures. The software would understand context, relationships, and the fluid nature of human thought. The default startup advice was to find a problem to solve—people buy drills because they want to make holes. But I was trying to build something to solve many different kinds of problems. The task felt more like trying to invent a new language—one that could express anything—rather than trying to make a drill. It was overwhelming. But I knew there were other people using these systems and probably doing it much more successfully than I was. Maybe if I interviewed them, if I talked to 50 top performers in different fields about what makes them tick, then I’d be able to derive a set of first principles, the physics of how knowledge organization worked. If I turned those interviews into a newsletter, I could also use the newsletter to build the audience for my eventual product. At the very least, it’d be an excuse to get interesting people to talk to me. The interviews resonated—tens of thousands of people subscribed to learn how others organized their thoughts and work. Eventually that newsletter, Superorganizers, grew into this company, Every. If you’re reading this now, maybe that’s how you first found my writing. What I didn’t expect from those interviews was learning that the experience I was going through—looking for the perfect organizational system—was common among the people I talked to. Many of them also had the same feeling that if they could just prepare enough, they could eliminate the uncertainty of failure. They’d say the right thing, or make the right decision at the right time—and that would make all the difference. I wasn’t the only one who wished I was a machine. Looking back, I realized there was a common thread to my attempts to find the perfect note-taking system, the perfect argument, the perfect philosophy. Those problems looked a lot like the problems we encountered when we tried to build an AI scheduling assistant at the beginning of this series: We can get by with a mechanical system of rules, but any system of rules can only capture a slice of reality. Still, I didn’t know of any alternative. I couldn’t name the metaphor I was operating under. And that’s why, almost as soon as I tried GPT-3 and began to learn how it worked, it took my breath away... Become a paid subscriber to Every to unlock this piece and learn about how: Every age models intelligence on its newest tools—from Plato’s wax tablets to today’s neural networks Language models break with 2,000 years of reductive thinking by embodying irreducible complexity New tools reveal what old ones hid, reshaping how we understand reality itself Upgrade to paid Click here to read the full post Want the full text of all articles in RSS? Become a subscriber, or learn more.

OpenAI is expanding Stargate to Michigan with a new one-gigawatt campus that strengthens America’s AI infrastructure. The project will create jobs, drive investment, and support economic growth across the Midwest.

OpenAI introduces Aardvark, an AI-powered security researcher that autonomously finds, validates, and helps fix software vulnerabilities at scale. The system is in private beta—sign up to join early testing.

A deep dive into OWL, the new architecture powering ChatGPT Atlas—decoupling Chromium, enabling fast startup, rich UI, and agentic browsing with ChatGPT.

OpenAI introduces gpt-oss-safeguard—open-weight reasoning models for safety classification that let developers apply and iterate on custom policies.

gpt-oss-safeguard-120b and gpt-oss-safeguard-20b are two open-weight reasoning models post-trained from the gpt-oss models and trained to reason from a provided policy in order to label content under that policy. In this report, we describe gpt-oss-safeguard’s capabilities and provide our baseline safety evaluations on the gpt-oss-safeguard models, using the underlying gpt-oss models as a baseline. For more information about the development and architecture of the underlying gpt-oss models, see the original gpt-oss model model card⁠.

DNP rolled out ChatGPT Enterprise across ten core departments, achieving 95% faster patent research, 10x processing volume, 87% automation, and 70% knowledge reuse in three months.

Discussion | Link

Researchers at Tsinghua University developed the Optical Feature Extraction Engine (OFE2), an optical engine that processes data at 12.5 GHz using light rather than electricity. Its integrated diffraction and data preparation modules enable unprecedented speed and efficiency for AI tasks. Demonstrations in imaging and trading showed improved accuracy, lower latency, and reduced power demand. This innovation pushes optical computing toward real-world, high-performance AI.

Doppel uses GPT-5 and reinforcement fine-tuning to stop deepfake and impersonation attacks, cutting analyst workloads by 80% and reducing response times from hours to minutes.

Microsoft and OpenAI sign a new agreement that strengthens its long-term partnership, expands innovation, and ensures responsible AI progress.

OpenAI’s recapitalization strengthens mission-focused governance, expanding resources to ensure AI benefits everyone while advancing innovation responsibly.

Discussion | Link

Meeting the demands of the Intelligence Age will require strategic investment in energy and infrastructure. OpenAI’s submission to the White House details how expanding capacity and workforce readiness can sustain U.S. leadership in AI and economic growth.

This system card details GPT-5’s improvements in handling sensitive conversations, including new benchmarks for emotional reliance, mental health, and jailbreak resistance.

OpenAI collaborated with 170+ mental health experts to improve ChatGPT’s ability to recognize distress, respond empathetically, and guide users toward real-world support—reducing unsafe responses by up to 80%. Learn how we’re making ChatGPT safer and more supportive in sensitive moments.

Steuerrecht.com uses ChatGPT Business to streamline legal workflows, automate tax research, and deliver faster, client-ready analysis for law firms.

HoloCine: Holistic Generation of Cinematic Multi-Shot Long Video Narratives