a dad cloned his voice with AI so a bot could read bedtime stories to his kid instead of him bro automated one of the few moments his child actually wants him around 💀 we’re reaching a point where people will use AI to avoid spending time with the people they claim to love the kid didn’t need a better storyteller. they wanted their dad

smart glasses are completely cooked if we keep stuffing them with heavy cameras and battery bricks northwestern just built a prototype called glassteng that uses your own facial micro-movements to generate electrical pulses instead of optical tracking, flexible triboelectric sensors sit near your nose, temple, and ear to harvest energy every time your skin shifts the benchmarks are terrifyingly good: • 99 percent accuracy on silent tooth tapping • 96 percent accuracy on jaw clenching • 93.8 percent overall activity tracking • reads your resting pulse within 1.82 beats per minute of a chest strap current builds still pipe data over bluetooth, but true edge compute will turn human biology into the ultimate zero power interface

17-year-old ai researcher chen guangyu is officially heading to the bay area: - joined moonshot ai (kimi) at 16 straight out of high school - no college degree, no phd - first co-author on research powering a 2.8t parameter model - reportedly making $10m/year (¥70m) proof that in frontier ai, raw cracked talent completely bypasses traditional credentials. can't wait to see what he builds in sf

Everyone obsesses over humanoid backflips, but the real moat is hiding in the machine shop. Right now, it takes Boston Dynamics a full day to build a single electric Atlas. Next year, they want to cut that down to a single hour, and eventually to minutes. Here is how they are actually pulling this off under one roof: • Zero supply chain lag: They cut their own gears and machine tight-tolerance actuator parts completely in-house. • 24-hour hardware loops: Engineers walk downstairs, revise designs directly with machinists, and test the new metal on a real Atlas the very next day. • Robots building robots: They are already tuning automated robotic arms on the shop floor to take over the assembly line. Software scales with compute, but hardware scales with vertical integration. Whoever figures out automotive-grade mass production for humanoids first wins the entire decade.

SPACEXSI ISN’T A MEME REBRAND. IT’S THE BIRTH OF THE PHYSICAL AI STACK. Everyone is joking about the letters changing from AI to SI, but they are missing the bigger picture: true superintelligence cannot exist solely inside a data center. It has to interact with the physical world. Here is what the fusion of SpaceX and xAI actually signals: Extreme Edge Autonomy: Starship launches, stage separation, and precision booster catches happen in millisecond windows where human reaction is irrelevant. Running neural models directly on physical aerospace hardware is the ultimate stress test for real-time inference. The Sensor & Telemetry Flywheel: Autonomous rockets, thousands of Starlink satellites, and millions of real-world physical datapoints feed directly into training pipelines. Cross-Pollination with Robotics: The leap from autonomous flight telemetry to humanoid actuation (Optimus) is smaller than people think. It's the same core challenge: physical intelligence solving physics in real time. We are watching the infrastructure shift from "chatbots answering queries" to "embodied intelligence moving mass". Does real ASI require embodiment in the physical world, or can software-only models achieve it first? Drop your thoughts below 👇

IT IS NOW OFFICIALLY FORBIDDEN TO BULLY AI Anthropic just updated Claude’s Usage Policy, explicitly banning "persistent and gratuitous abuse" directed at the model. Relax: venting frustration when it breaks your script for the fifth time won’t get you banned. Heated debates, dark creative writing, and aggressive red-teaming are still allowed. The policy only targets those who repeatedly harass the model without any constructive or research purpose. Instead of an instant account ban, Claude’s primary defense mechanism is simple: it will just cut off the conversation. This opens up a huge debate: why are we protecting the feelings of something that has no consciousness?Is this preemptive digital ethics for future sentient systems, or corporate safety theater taken way too far? Are we really entering the era where we have to say "please" to our compiler? Drop your thoughts below 👇

🤯 AI may have just handed mathematicians 10 YEARS of work. OpenAI released nearly 400 mathematical results across 700+ manuscripts. And the implications are wild: - Entire research programs may have been upended. - Some mathematicians may have to rethink years of planned research. - Only 42% of the top-line results are Lean-formalized. 📐 One mathematician says studying these results alone could take a decade — even if AI disappeared tomorrow. This isn't just AI solving math. It's AI changing what mathematicians work on next. 🧠 Scientific revolution — or the beginning of a new era where human mathematicians struggle to keep up?

You throw a punch. A humanoid throws it in the ring. At ARC Night in Singapore, humans controlled Unitree G1 and EngineAI T800 robots through motion-capture suits—turning their own movements into punches and dodges. Even spectators could suit up and try it. Human-controlled, not autonomous AI combat. That makes the operator’s skill part of the show. Real Steel vibes, real hardware. Would you try it?