3297626
A hub for startup news, trends, and insights, covering the global startup ecosystem for founders, investors, and innovators. Community: @startupdis Buy Ads: @strategy (this is our only account).
🧱 LEGO teaches bricks to respond to play
At CES 2026, LEGO introduced SMART Play, a new system the company calls its biggest leap since the minifigure era, physical blocks that react to what kids actually do with them, without screens or apps.
🔸 A responsive building system: SMART Play is made up of connected bricks, tags, and figures that give traditional LEGO builds awareness and feedback.
🔸 Electronics hidden in plastic: A regular-looking 2×4 brick contains a custom chip, motion sensors, LEDs, a speaker, and wireless charging components.
🔸 Meaning, not scripts: Small ID tiles and minifigures tell the system what a build represents, allowing behavior and sound to change as the story evolves.
🔸 Local intelligence: Using a Bluetooth-based network, bricks understand position, movement, and orientation relative to each other in real time.
🔸 Unlimited sound design: Audio is generated on the fly rather than stored, so the same brick can convincingly sound like engines, animals, or something far less heroic.
🔸 Long-term bet: The platform took around a decade to develop and is protected by dozens of patents.
🔸 First release: The system launches with Star Wars sets priced from $70 to $160, with pre-orders opening January 9 and shipping starting March 1.
SMART Play doesn’t pull LEGO into screens, it quietly adds intelligence to the bricks themselves, keeping play tactile while making it adaptive.
📊 @tech
🗣️Geoffrey Hinton on why cancer could stop being a death sentence
Most deaths happen because tumors are found too late, not because medicine lacks treatments. Full-body MRI can catch changes early. The issue is scale.
Humans can’t read that much data consistently, AI can.
📊 @tech
🖼 AI-generated art moves to the wall
At CES 2026, Fraimic unveiled Smart Canvas, an E-ink display in a picture frame that replaces traditional wall art with AI-generated images you can change by voice.
🔸 The Smart Canvas uses E-ink Spectra 6, designed to mimic the texture of real canvas and paint rather than a glossy screen, making it look closer to physical art than a digital display.
🔸 Images can be generated via voice commands, with a built-in AI powered by OpenAI, or uploaded manually through a web interface if you want full control over the visuals.
🔸 Power consumption is minimal: the battery is rated to last several years, since E-ink only draws power when the image changes.
🔸 The panel can be placed into any standard frame and mounted in any orientation, making it feel more like a modular art object than a gadget.
🔸 Two versions are available: a standard size for $400 and a large version for $1,000. Pre-orders are live, with shipping expected to start in June.
🔸 The concept works best for abstracts, patterns, and generative art, though there’s a real risk that day-to-day use exposes familiar AI flaws: warped lines, odd artifacts, and “same-y” compositions.
Smart frames won’t replace real art, but they hint at a new category where walls become programmable and this market is only just beginning to take shape.
🔥 An AI streamer just beat every human on Twitch
Neuro-sama, an AI VTuber created by programmer Vedal, is now the most subscribed streamer on the platform. Around 162,000 active subs. Second place isn’t even close.
Neuro-sama runs near 24/7, chats with viewers, sings, plays games, reacts to videos, and is powered by multiple custom AI systems, not a single prompt. At standard Twitch splits, subscriptions alone likely bring in over $400,000 per month.
Ads, donations, and sponsorships are on top.
An AI personality outperformed the biggest human creators. Streaming just crossed a line.
📊 @tech
⏱️ Pebble brings back the round watch.
The revived Pebble brand keeps leaning into what made it special. After Pebble 2 Duo and Pebble Time 2, the round model is back with Pebble Round 2.
🔸 The original Time Round had one big flaw: a thick bezel around a tiny 1-inch screen. The new version fixes that completely. A 1.3-inch display now fills the entire front, with resolution doubled to 260 × 260.
🔸 Battery life is the real flex. Up to 2 weeks on a charge, thanks to a color e-paper display, the same tech used in e-readers.
🔸 The case is stainless steel, available in black, silver, and rose gold. Inside are 2 microphones for voice input. You can talk to AI assistants or дикtate replies to messages, with full support on Android for now.
🔸 This is not a sports watch, and Pebble is clear about that. It tracks steps and sleep, but serious training is not the goal.
It’s a simple smart watch with character, built for people who miss that Pebble vibe.
Preorders are open at $200. Shipping starts in May.
📊 @tech
🗣️Sam Altman: “Most people don’t take enough risk”
“I think people have terrible risk calculus in general… almost always A) you’re wrong about what is risky and what is not risky, and B) most people don’t take enough risk especially early in your career. Being young, unknown, and poor, is actually a great gift in terms of the amount of risk you can take.”
“I think what risk actually looks like is not doing something that you will spend the rest of your life regretting… So if you really believe in something if there’s an idea you’re super passionate about and you take a calculated risk to start a company realizing you may forego a couple of years of steady income and maybe people call you a failure, that’s a great risk to take. And if you don’t take that risk, I think you have a very high chance that you end up regretting that.”
“One really important thing to strive for in your career is to be a doer, not a talker. And the reason that people don’t do stuff is 1) it’s hard, and 2) it’s risky. And so you have these people that want to dabble in a bunch of different projects, but never be all-in on one… I think that’s really bad. I think history belongs to the doers, and I think you should take a risk and actually do something.”
Google's CEO gets asked, "Does Google track you through your phone?"
📊 @tech
🔥 These robots have moved a building!
A team of 432 walking robots is carefully moving a 7,500-ton historic building in Shanghai. Instead of traditional machinery, these robots gently lift and “walk” the building about 10 meters per day.
The area is densely packed with narrow alleys and old structures, making cranes and large machines unusable. These robots were chosen because they can operate in tight spaces and move precisely without damaging nearby buildings.
In China robots are even moving existing buildings! 🤯
📊 @tech
Happy New Year!
Thanks for reading and staying curious.
See you in 2026.
📊 @tech
🔥 A 10-hour fireplace turned into a full-time income
There’s a YouTube channel with a single video: ten hours of a burning fireplace. No edits, no commentary, no updates.
That one video brings its creator around $180k usd a year.
Sometimes the smartest business model is realizing people just want something calm playing in the background.
📊 Powered by Crypto Insider
When AI finally runs the world, you’ll know exactly where to point the finger.
P.S. GPUs owe their existence to games like Quake and Half-Life 2. Even Demis Hassabis, now leading AI at Google, began as a game developer.
Turns out the gaming industry wasn’t just entertainment, it quietly laid the groundwork for modern AI, and its role is still massively underestimated.
📊 Powered by Crypto Insider
🔥 Claude Code is now writing 100% of its own code and its creator let it happen
Claude Code’s lead developer has admitted that every single contribution over the past 30 days was written by Claude itself not as an experiment, but as normal workflow.
🔸 The developer is Boris Cherny, widely regarded as the original creator and “founder” of Claude Code, which began as his side project in 2024.
🔸 In the last month alone, Claude Code:
• shipped 259 pull requests
• made 497 commits
• added ~40,000 lines of code
• deleted ~38,000 lines, with every line generated by Claude Code + Opus 4.5.
🔸 The human role has shifted to orchestration: prompts, reviews, architecture decisions not typing code.
🔸 Claude can now run continuously for minutes, hours, even days, using Stop hooks to manage long-lived development loops.
🔸 This isn’t toy automation or autocomplete, it’s end-to-end software production at scale, driven by the system itself.
🔸 What started as a side project has quietly become a proof point for a new programming paradigm.
Software development is crossing from human-written with AI assistance to AI-written with human supervision and this is likely the slowest this transition will ever be.
📺 LEGO TV recreates the hidden guts of a 1960s CRT set
A LEGO enthusiast has gone beyond retro styling and rebuilt what actually lived behind the back panel of old televisions, down to the physics.
🔸 Creator FMDavid designed a LEGO TV that isn’t just a shell, but a full internal reconstruction inspired by real 1960s cathode-ray televisions.
🔸 The model features a large CRT tube dominating the interior, with surrounding deflection coils that, in real TVs, steered the electron beam across the screen.
🔸 Inside are LEGO versions of vacuum tubes, capacitors, transformers with cooling fins, and detailed color wiring between components.
🔸 The back panel even includes RCA connectors, reinforcing the illusion that this is a functional analog device.
🔸 The screen is convex, not flat a subtle but accurate detail, since real CRTs needed curved glass to withstand atmospheric pressure against the vacuum inside.
🔸 The project is now live on LEGO Ideas; if it reaches 10,000 votes, LEGO may consider turning it into an official set.
This isn’t just nostalgia, it’s a reminder that past tech was physical, visible, and mechanical, and that understanding how things worked can be just as compelling as sleek modern design.
When the robot rebellion begins, no one will know, but there will be signs
📊 Powered by Crypto Insider
🐬 Disney’s quiet bet on aquatic robots
Disney has been developing robotic aquatic creatures for years and what started as simple underwater tech is now starting to look like a full replacement for live-animal shows.
🔸 Disney’s work on aquatic robots began over 20 years ago, initially as basic underwater drones with motors for vertical and horizontal movement closer to radio-controlled vehicles than lifelike robots.
🔸 Around 2020, Disney shifted toward biomimetic designs, unveiling dolphin-inspired robots that used fins instead of propellers, dramatically improving fluidity and realism.
🔸 By 2025, Disney introduced near-autonomous hydrofoil-based aquatic robots equipped with GPS, ultrasonic sensors, and self-balancing systems.
🔸 These robots are currently designed for live shows inside Disney parks, where predictability, safety, and choreography matter more than full autonomy.
🔸 Disney says the designs are inspired not only by real marine animals, but also by fictional creatures from Avatar: The Way of Water, blending biology with cinematic imagination.
🔸 The implication is hard to miss: realistic robotic animals offer spectacle without ethical backlash, training costs, or regulatory pressure.
What you’re seeing isn’t just a theme-park experiment, it’s an early glimpse of a future where dolphinariums are replaced by code, sensors, and storytelling instead of live animals.
✅ Boston Dynamics unveils Atlas, now built for real factory work
This isn’t a flashy lab demo anymore. Atlas has crossed the line from stunt videos to industrial hardware meant to replace manual labor on the factory floor.
What makes it different:
• Atlas recharges itself: it walks to the station, removes a depleted battery, inserts a fresh one, and keeps going. No downtime, no breaks, 24/7 operation.
• AI inside: Boston Dynamics is working with Google DeepMind, bringing neural networks into Atlas so it can reason, adapt, and learn new tasks instead of following rigid scripts.
Key specs:
• Lifts up to 50 kg
• Height: 2.3 meters
• 56 degrees of freedom, enabling human-like (and sometimes inhuman) movement
• Resistant to water and cold, ready for harsh industrial environments
Production plans:
• Serial assembly has already started in Boston
• All 2026 deliveries are booked, first units go to Hyundai factories and Google DeepMind
• Wider availability pushed to 2027
• A dedicated factory is planned with capacity for 30,000 units per year
This isn’t a robot for demos, it’s a shift in how factories are staffed. And no, Atlas won’t work for $300 a month… even robots have standards now.
📊 @tech
🔔 An EV startup thinks it can sell a $25k truck by doing the opposite of Tesla
Slate Auto plans to ship a $25,000 electric truck in the US by end of 2026, without tax credits.
The core idea:
• 1 model, 1 configuration
• 600 parts instead of 2,500
• Composite body panels, no paint shop or stamping
• No built-in infotainment, phone or tablet instead
• Manual windows, AC included
Everything else follows from that. Simplicity keeps costs down, while customization and accessories provide margin. Slate isn’t chasing Tesla buyers. It’s targeting people choosing between a $27k used car and a new vehicle with a warranty.
The bet is simple: Reduce complexity, lower capital costs, and reach profitability earlier by building less.
📊 @tech
🤖 UBTECH showed Walker S2 playing tennis
It’s a highlight-style demo, but still notable. The humanoid tracks the ball, adjusts its footwork, and returns shots in real time.
Walker S2 is expected to enter mass production this year.
📊 @tech
🔔 Wild tweet by Elon Musk
For anyone not deep in AI lore, the singularity means a point where AI improves itself faster than humans can understand or control.
Systems start designing better versions of themselves. That’s when progress stops checking who’s in the room.
📊 @tech
🎥 Missed DeepMind’s The Thinking Game?
The documentary tracing DeepMind’s journey is now available to watch at no cost.
It premiered back in 2024, but the full version was released publicly in late November 2025 and quickly went viral, racking up 200M views on YouTube in just four weeks.
👉 Watch here 👈
📊 @tech
✅ A BlackBerry-style phone makes a surprising comeback
Clicks. known for physical keyboard phone cases has just unveiled its first standalone smartphone built around a full QWERTY keyboard and minimalist communication focus.
🔸 Physical keyboard at the center: The Communicator features a full QWERTY keyboard with larger keys and touch-sensitive functionality that can also act as a trackpad aiming to make typing and navigation faster and more tactile than on a typical touchscreen.
🔸 Standalone 5G Android device: It isn’t just an accessory, the Communicator has its own cellular connection, runs Android 16 with a stripped-back launcher to reduce distraction, and supports essential apps like Gmail, WhatsApp, Slack, and Telegram.
🔸 Retro-modern hardware: The phone packs a 4.03-inch OLED screen, front and rear cameras, a 4,000 mAh battery, a headphone jack, expandable microSD storage, and interchangeable back plates for personalization.
🔸 Design with everyday features: Clicks has included a physical mute switch and a customizable side LED key for notifications or shortcuts, little touches that mainstream phones have dropped.
🔸 Price and preorder: The Communicator is priced at about $499, but early reservations (with a $199 deposit) drop it to $399 ahead of general availability later this year.
🔸 Focus on communication, not consumption: With a minimalist interface and hardware built for typing and messaging, the phone targets users who want a less distracting alternative to traditional smartphones.
Clicks is betting that tactile control and focus-oriented design still matter in a world dominated by large, distraction-driven touchscreen devices.
Ai is mysteriously learning things it was never programmed to know.
📊 @tech
The money talk every parent should give their kids.
📊 @tech
🎓 MIT drops a free course on AI agents, theory and real-world practice
The folks at MIT have released a free, comprehensive course on AI agents, covering both fundamentals and applied use cases.
🔸 AI agents, what they are, how they differ from chatbots, and where they actually make sense.
🔸 Planning, task decomposition, decision-making loops, and agent control logic.
🔸 RAG (Retrieval-Augmented Generation), grounding agents in external knowledge and tools.
🔸 MCP (Model Context Protocol), structuring context, memory, and tool access for scalable agents.
🔸 Examples & case studies, real implementations, not just academic theory.
🔸 Designed as a practical base for anyone building agentic systems in production.
MIT isn’t just teaching how to talk to models, it’s teaching how to make AI act, which is quickly becoming the most valuable skill in applied AI.
🔥 A helmet straight out of Cyberpunk
Looks like gear worn by NPCs in Night City but it’s real. This is Alpha Wave, a futuristic bicycle helmet from Also, and it’s packed with unconventional safety tech.
🔸 The core innovation is the Release Layer System, an alternative to classic MIPS.
🔸 Under the outer shell panels sits a layer of tiny polycarbonate balls.
🔸 On impact, a special adhesive releases the shell panel at the contact point, allowing it to roll freely in any direction and then detach.
🔸 This movement dissipates rotational energy, the kind that usually gets transferred directly to the brain.
🔸 The chin strap is redesigned too. The HighBar system replaces the usual Y-straps with rigid mounts, flexible straps, and a one-handed adjustment wheel under the chin, usable even with gloves.
🔸 Lighting is built in:
• 200-lumen headlight up front
• 75-lumen work light for digging through a backpack in the dark
• Rear light wirelessly syncs with Also’s TM-B electric bike
🔸 Audio is fully integrated:
• 4 speakers with wind protection
• 2 noise-canceling microphones for calls and navigation prompts
🔸 Specs: IPX6 water resistance, USB-C charging.
🔸 Price: $250, not cheap, especially since a serious impact means full replacement.
🔸 Pre-orders open soon, with delivery starting early next year.
It’s expensive, disposable after a crash, and unapologetically futuristic but when it comes to your head, everyone decides for themselves what protection is worth.
⚡️ China’s superconducting maglev hits 700 km/h in a 400-meter sprint
Chinese engineers from the National University of Defense Technology (NUDT) accelerated a monocoque test vehicle to 700 km/h and back to zero on just a 400-meter track, setting a new benchmark for superconducting maglev systems.
🔸 The acceleration reached ~10 g, well beyond what even trained fighter pilots can tolerate; braking peaked at around 5 g, confirming this is a tech demo, not a passenger vehicle.
🔸 The system uses high-temperature superconductors, cooled with liquid nitrogen to −196°C far simpler and cheaper than classic liquid-helium superconductors at −269°C.
🔸 This makes the setup more practical for real-world infrastructure, while still enabling extreme speeds and stability.
🔸 The stated next milestone is 1,000 km/h, faster than the cruising speed of commercial passenger aircraft.
🔸 At that speed, a Shanghai–Beijing trip could theoretically drop to ~2 hours, versus ~14 hours by car today.
🔸 China already operates the world’s only commercial high-speed maglev: the Shanghai Maglev, which covers 30 km in 7–8 minutes at up to 431 km/h.
China isn’t just experimenting with futuristic transport, it’s systematically pushing maglev toward air-travel territory, where trains start competing with planes rather than cars.
🔒 Google released a mega-guide with 40 tips for working with modern AI models
Google compiled a practical playbook for using today’s most hyped neural networks, essentially a year’s worth of AI learnings packed into one guide.
🔸 The guide includes non-obvious Gemini 3 features, showing where it outperforms expectations and how to unlock hidden capabilities.
🔸 It collects AI tools, prompt patterns, and workflows that go beyond basic chat usage aimed at real productivity gains.
🔸 Many tips focus on combining models and tools, not treating any single AI as a silver bullet.
🔸 The recommendations are task-agnostic: writing, coding, research, planning, analysis, and creative work all fit the framework.
🔸 A recurring theme is leverage how small prompt or setup changes can lead to disproportionately better outputs.
🔸 The guide positions AI less as a chatbot and more as a programmable co-worker that rewards structured thinking.
AI skill is no longer about knowing which model is best, it’s about knowing how to extract leverage from any of them.
🧠 Andrey Karpathy says he feels like a lagging programmer in the AI era
Former OpenAI and Tesla AI leader Andrey Karpathy shared a candid reflection on how AI is radically reshaping what it means to be a programmer and why even top engineers now feel behind.
🔸 Karpathy says human contribution to code is becoming sparse and fragmented, with large parts delegated to AI systems rather than written directly.
🔸 He believes he could be 10× more productive if he fully mastered the new AI tooling stack that has emerged just in the last year and calls his inability to do so a skill issue.
🔸 A new abstraction layer has appeared on top of traditional software engineering: agents and subagents, prompts, contexts, memory, modes, tools, plugins, skills, hooks, MCP, LSP, slash commands, IDE integrations, and workflows.
🔸 Developers now need a mental model of stochastic, error-prone, constantly changing systems very different from deterministic software yet deeply intertwined with classic engineering.
🔸 Karpathy compares AI to a powerful alien tool with no manual, where everyone is forced to experiment in real time while the profession undergoes a magnitude-9 earthquake.
🔸 His advice is blunt: roll up your sleeves, because the pace of change means standing still is falling behind.
Karpathy’s post reads less like a complaint and more like a warning: programming isn’t dying but the old mental models are, and even the best engineers now have to relearn the craft from scratch.
🧠 A 33,000-skill “brain” for AI agents is emerging
A massive, structured skill database is being built for Claude Code, OpenAI Codex, and ChatGPT, allowing neural networks to be explicitly taught new capabilities instead of relying on ad-hoc prompts.
🔸 The database contains 33,000+ distinct skills spanning automation, software development, research, data workflows, and operational tasks.
🔸 Each skill comes with clear instructions, constraints, and expected behavior, making them reusable rather than one-off prompt hacks.
🔸 A search engine and filters make it possible to quickly find the right capability by domain or use case.
🔸 Skills are linked to real sources on GitHub and documentation, grounding them in practical implementations.
🔸 The system is designed to plug directly into agentic workflows for coding agents and autonomous AI systems.
This marks a shift from prompt engineering to capability engineering where AI systems are upgraded by importing skills, not rewriting prompts.