Stanford researchers found that people produced about 60% more creative ideas while walking than while sitting still. Which means the single most reliable creativity intervention ever measured is free, requires no training, and is structurally forbidden by the way almost every office, classroom, and Zoom call on earth is arranged. The study came out of Stanford in 2014. The lead researcher was Marily Oppezzo, then a doctoral student in educational psychology, working with her advisor Daniel Schwartz, a professor at the Stanford Graduate School of Education. Here is the detail that sounds invented and isn't. They had the idea for the study while they were out on a walk together. The paper is called "Give Your Ideas Some Legs: The Positive Effect of Walking on Creative Thinking." It appeared online in April 2014 in the Journal of Experimental Psychology: Learning, Memory, and Cognition. Four experiments. 176 participants, mostly college students. The design was within-subject, which matters more than it sounds: they weren't comparing walkers to sitters, they were comparing the same person to themselves. Same brain, same day, seated and then moving. To measure creativity they used Guilford's Alternate Uses test, a standard instrument that asks you to name novel but appropriate uses for an everyday object. Given a button, you might say a doll's eye, a tiny strainer, a doorknob for a dollhouse. Two conditions had to be met for an idea to count. It had to be novel, meaning nobody else in the group had said it. And it had to actually be feasible. "A button as a spaceship" earns you nothing. Then the numbers. Across the four experiments, the proportion of participants who were more creative walking than sitting was 81%, then 88%, then 95%. And in one of them, 100%. Every single person tested. Not a trend, not a modest signal buried in error bars. Everybody. Now, if you have any scientific instinct at all, you are already objecting. And your objections are the right ones, which is why the study is worth your attention rather than a passing nod. Obvious objection one: it isn't the walking. It's the outdoors. Sunlight, air, trees, birds, the visual richness of a world that isn't a monitor. Of course people have better ideas out there. So they put people on a treadmill. Indoors. In a small room. Facing a blank wall. No scenery. No fresh air. No novelty. Nothing to look at but paint. The effect held. Oppezzo has said she expected outdoor walking to be the dramatic condition, the one that would blow the others away, and that she was genuinely surprised the tedious blank-wall treadmill worked as well as it did. Which is the reaction of someone whose own hypothesis got beaten by her control. Obvious objection two, the better one: it still isn't the legs. It's the motion. Being moved through space, watching the world slide past you, the gentle physical rhythm of going somewhere. Any of that could do it. So in the fourth experiment they took people outdoors and split them. Some walked. Some were pushed through the same environment in a wheelchair. Sit with that design for a second, because it's beautiful. Both groups are outside. Both see the same trees, the same path, the same sky. Both are physically moving through space at roughly a stroll's pace. The scenery is matched. The motion is matched. The novelty is matched. The only difference is whether your own legs are doing the work. The walkers were substantially more creative. That is what isolates the effect. Not the air, not the view, not the going-somewhere. The act of walking. Here is the part that pop-science summaries almost always leave out, and it's the part that makes the finding usable instead of merely inspiring. Walking did not improve all thinking. On convergent tasks it made performance slightly worse. Convergent thinking is the single-right-answer kind. The study used compound remote associates: you get three words and have to find the one word that links them all. Cottage, Swiss, cake. The answer is cheese. There is exactly one answer and you either close on it or you don't. Walkers were a bit worse at that than sitters. So the honest version of the finding is not "walking makes you smarter." It's that walking widens the aperture. It multiplies possibilities and loosens associations, and it does that at some cost to narrow, convergent focus. Which gives you a rule that fits in six words. Walk to generate. Sit to decide. Then there's the finding I'd argue is the most practically valuable in the whole paper, and it gets almost no airtime. The effect lingered. Participants who walked and then sat down were still measurably more creative during the seated session afterward. You do not need a walking desk. You do not need to take calls while pacing or dictate ideas into your phone mid-stride. You can walk, then come back and work. The residue is the point. As for why any of this happens, the honest answer is that this paper didn't establish it. Schwartz said plainly that the causal mechanisms still needed to be worked out. Later work in cognitive neuroscience has offered plausible accounts, usually involving mind-wandering and the brain networks associated with it, the idea being that mild rhythmic movement quiets deliberate control and lets loose association run. That's a reasonable story. It is not something these four experiments demonstrated, and anyone telling you it is has read the headline and not the paper. Oppezzo's own framing was appropriately modest. Nobody is claiming walking turns you into Michelangelo. It helps at the front end of the creative process, the stage where you are still generating raw material. What's strange is how thoroughly people had already figured this out without the data. Darwin built a gravel path behind his house, the Sandwalk, and looped it daily as his thinking route. Nietzsche walked enormous distances and was convinced his best ideas arrived only on foot, that thinking done sitting down was somehow suspect. Beethoven took long afternoon walks carrying pencil and paper, on the assumption that something worth writing down would show up. Kahneman and Tversky, who between them rebuilt our understanding of human judgment, did much of their best collaborative thinking on unhurried walks together. Steve Jobs conducted meetings on foot as a matter of habit, a practice other tech founders have since copied. None of them had 176 participants or a wheelchair control. They just noticed what happened to their own minds and organized their lives around it. The study didn't discover the effect. It confirmed what a long line of people had been quietly exploiting for centuries. So consider every seated brainstorm you have ever sat through. The whiteboard, the shared doc, the polite silence while everyone waits for someone else to have the idea. Consider the last hour you spent stuck on a problem, staring at the same screen, certain the answer would arrive if you just stayed put a little longer. It was probably in your legs. No app. No subscription. No cold plunge, no supplement stack, no productivity system with a name. Fifteen minutes and a pair of shoes. The chair is not where ideas come from.

Microsoft open-sourced a 4B model that turns any image into a production-ready 3D asset in 3 seconds. It’s called TRELLIS.2, a fully textured, physically accurate 3D models with PBR textures out of the box. → Full PBR (base color, roughness, metallic, opacity) → Handles hair, cloth, glass, non-manifold geometry → Exports .glb ready for Unity/Unreal/Blender → Runs locally, ships in 3 seconds It's not a demo or a research preview. The full training codebase is public. You can fine-tune it on your own asset library and get a model that generates in your studio's exact style. 100% Open Source

Microsoft open-sourced a 4B model that turns any image into a production-ready 3D asset in 3 seconds. It’s called TRELLIS.2, a fully textured, physically accurate 3D models with PBR textures out of the box. → Full PBR (base color, roughness, metallic, opacity) → Handles hair, cloth, glass, non-manifold geometry → Exports .glb ready for Unity/Unreal/Blender → Runs locally, ships in 3 seconds It's not a demo or a research preview. The full training codebase is public. You can fine-tune it on your own asset library and get a model that generates in your studio's exact style. 100% Open Source

Why does Claude search the web up to 20 times just to answer one simple question? Say you ask for the best noise-cancellation headphones under $200 released this year. First, Claude figures out what you're actually asking for. In this case, it requires real-time data. So, Claude decides to use the web search tool because Anthropic explicitly instructs it: "If you see something that requires up-to-date information, you must use the web search tool." Claude converts your prompt into something short and specific, three to four words, like best headphones 2026. Then Claude's infrastructure makes the web API call. Anthropic uses the Brave API for searching the web because Brave hands back more than just links. It gives you titles, snippets, and rich metadata. Once it has the first 10 search results, it runs a ReAct loop: reason then act on repeat. It reads what came back and picks one of three moves: • Narrow the search query and try a different search. • Open the full page of one of the search results to get more details. • Decide: "I have enough information to answer the question." Along the way, it’s constantly checking sources. A well-known review site is going to outrank a random blog. Then it loops again and again, sometimes up to 20 times, until every part of the answer is actually backed by something it found. And if we add up all the costs from the Brave API to the ReAct loop, on average, a single deep web search session costs Anthropic anywhere between 8 to 25 cents. We think of AI as a single response. Behind the scenes, it’s an army of search queries burning cash to make sure it doesn't hallucinate.

Someone open sourced a full 3D flight simulator that runs entirely in your browser.. real-world terrain. real locations. zero downloads. you can literally fly a plane or drive a car anywhere on earth for completely free

In 2013, Facebook bought an Israeli VPN app for $120M and turned it into a surveillance tool that spied on 33M+ users entire phones for years. This helped Zuck buy WhatsApp for a whopping $19B and break Snapchat's encryption. The app was called Onavo Protect. It was sold as a free VPN that would secure your data and cut your mobile data usage. The justification was mundane and technically true. A VPN routes traffic through a remote server, and the server it routed through belonged to Facebook. In reality the routing was the product. Every request leaving the handset passed through infrastructure Facebook controlled, and the company kept a record of what passed through it. Captured: - apps installed on the device - session length per app - websites visited, with timestamps - volume of data moved This was not market research. It was an early-warning radar for competitors. Onavo data reportedly showed WhatsApp carrying messaging volume far above Messenger's. In February 2014, Facebook announced it was acquiring WhatsApp for $19 billion. There was just one catch. Facebook had offered roughly $3 billion in cash for Snapchat in November 2013 and been turned down by Evan Spiegel, then 23; nearly three years later Snapchat's traffic was encrypted, and Onavo could see that data was moving but not what it was. On 9 June 2016 Mark Zuckerberg wrote to his executives that the company needed "a new way to get reliable analytics about them." The effort was codenamed Project Ghostbusters, after the ghost in Snapchat's logo, and it sat inside a program called the In-App Action Panel. About a month later, in July 2016, Onavo engineers proposed kits installed on iOS and Android that would intercept traffic for specific subdomains, described in the email itself as a man-in-the-middle approach. The kit installs its own root certificate on the handset, so the phone trusts the interceptor and hands over the request in the clear. Nothing is broken. The trust model is simply pointed somewhere else. This ran on a panel of selected devices, not across all 33 million installs, and the technique was later extended to YouTube from 2017 and to Amazon from 2018, while Onavo's internal charts on rival app adoption sat among the documents released by the UK Parliament's DCMS committee in 2018. In August 2018 Apple removed Onavo Protect from the App Store for violating its data-collection rules, and Google Play pulled it in February 2019. Running alongside Onavo since 2016 was a second program, Facebook Research, which paid participants aged 13 to 35 between $10 and $20 a month for root-level access to their phone's network traffic. Its code contained numerous references to Onavo. TechCrunch published it on 29 January 2019. Apple revoked Facebook's enterprise developer certificate the following day, breaking the company's internal iOS apps worldwide. Australia's competition regulator separately sued Facebook over the representations Onavo Protect made between February 2016 and October 2017. Plaintiffs in the antitrust case that unsealed these documents in March 2024 did not describe the program as merely anticompetitive. They described it as criminal, and alleged Facebook's own lawyers were closely involved in designing and expanding it. Some of the company's senior engineers had objected internally. A privacy product is a collection point with better branding. It was downloaded 33 million times by people who wanted their data protected. They got their data measured, and Facebook got a list of what to buy or clone.

In 2013, Facebook bought an Israeli VPN app for $120M and turned it into a surveillance tool that spied on 33M+ users entire phones for years. This helped Zuck buy WhatsApp for a whopping $19B and break Snapchat's encryption. The app was called Onavo Protect. It was sold as a free VPN that would secure your data and cut your mobile data usage. The justification was mundane and technically true. A VPN routes traffic through a remote server, and the server it routed through belonged to Facebook. In reality the routing was the product. Every request leaving the handset passed through infrastructure Facebook controlled, and the company kept a record of what passed through it. Captured: - apps installed on the device - session length per app - websites visited, with timestamps - volume of data moved This was not market research. It was an early-warning radar for competitors. Onavo data reportedly showed WhatsApp carrying messaging volume far above Messenger's. In February 2014, Facebook announced it was acquiring WhatsApp for $19 billion. There was just one catch. Facebook had offered roughly $3 billion in cash for Snapchat in November 2013 and been turned down by Evan Spiegel, then 23; nearly three years later Snapchat's traffic was encrypted, and Onavo could see that data was moving but not what it was. On 9 June 2016 Mark Zuckerberg wrote to his executives that the company needed "a new way to get reliable analytics about them." The effort was codenamed Project Ghostbusters, after the ghost in Snapchat's logo, and it sat inside a program called the In-App Action Panel. About a month later, in July 2016, Onavo engineers proposed kits installed on iOS and Android that would intercept traffic for specific subdomains, described in the email itself as a man-in-the-middle approach. The kit installs its own root certificate on the handset, so the phone trusts the interceptor and hands over the request in the clear. Nothing is broken. The trust model is simply pointed somewhere else. This ran on a panel of selected devices, not across all 33 million installs, and the technique was later extended to YouTube from 2017 and to Amazon from 2018, while Onavo's internal charts on rival app adoption sat among the documents released by the UK Parliament's DCMS committee in 2018. In August 2018 Apple removed Onavo Protect from the App Store for violating its data-collection rules, and Google Play pulled it in February 2019. Running alongside Onavo since 2016 was a second program, Facebook Research, which paid participants aged 13 to 35 between $10 and $20 a month for root-level access to their phone's network traffic. Its code contained numerous references to Onavo. TechCrunch published it on 29 January 2019. Apple revoked Facebook's enterprise developer certificate the following day, breaking the company's internal iOS apps worldwide. Australia's competition regulator separately sued Facebook over the representations Onavo Protect made between February 2016 and October 2017. Plaintiffs in the antitrust case that unsealed these documents in March 2024 did not describe the program as merely anticompetitive. They described it as criminal, and alleged Facebook's own lawyers were closely involved in designing and expanding it. Some of the company's senior engineers had objected internally. A privacy product is a collection point with better branding. It was downloaded 33 million times by people who wanted their data protected. They got their data measured, and Facebook got a list of what to buy or clone.

Someone open sourced a full 3D flight simulator that runs entirely in your browser.. real-world terrain. real locations. zero downloads. you can literally fly a plane or drive a car anywhere on earth for completely free

Researchers proved every major LLM is secretly obsessed with Japan. And they finally figured out why. For years, we’ve been told that AI is entirely Western-centric, that it just reflects Silicon Valley and American values. A landmark paper by Cardiff and Basque researchers tested 31,680 cultural prompts across 24 languages on frontier models like ChatGPT, Claude, and Gemini. The results shattered that assumption. In six out of eight frontier models, Japan was the single most frequently referenced country when asked open-ended cultural questions. Ask about traditional dances, festivals, or everyday practices in an open context, and the AI defaults to Japan. Over and over again. Here is the twist nobody expected. This bias doesn't come from raw pre-training internet data. The researchers tracked where the obsession forms. It emerges after pre-training, during the supervised fine-tuning and alignment phase when humans teach the AI how to behave. Why Japan? Because decades of global soft power, rich cultural export, and clean, universally admired digital archives make Japanese culture uniquely "safe" for AI safety filters to lean on. When labs train models to be harmless and universally pleasing, the AI defaults to the cultural equivalent of comfort food. It avoids controversy by talking about anime, sushi, and tradition.

NVIDIA open-sourced a 600M model that transcribes 40 languages in real-time at 80ms latency and it costs $0. that's faster than you can blink. across mandarin, arabic, hindi, portuguese, tagalog, whatever,from a SINGLE checkpoint. → 17x more concurrent streams than buffered ASR on the same H100. → punctuation + capitalization built-in. no post-processing. → runs on your own GPU. no API bill 100% Open Source.