Everyone says AI kills jobs. Small businesses using AI plan to hire MORE than those that don’t. Does this match what you’re seeing? 👇

Today @RealbotixCorp is live with its 1st commercial robot line. Check out our new website Realbotix.com and reserve yours today.

💡 AI tip of the day Stop asking AI for answers. Make it ask you questions first. Add this to any prompt: “Before you answer, ask me the 5 questions that would most change your answer. Wait for my replies.” Why it works: AI fills gaps with generic assumptions. This forces it to do what a good PM does: gather requirements before building. Generic in = generic out. Questions first = answers that fit. What’s your go-to prompt add-on? 👇

Crimson Desert DLC vs. Assassin’s Creed Black Flag Resynced Sailing Comparison

📊 AI stat of the day 12 months ago ChatGPT had 72% of AI web traffic. Now it’s 56%. Gemini nearly doubled. Claude grew 4x. What it means: AI is no longer a one-app market. Users shop around. Why it matters for builders: don’t lock your product, prompts or SEO strategy to a single model. Build for a multi-AI world.

Let’s connect 🤝 I post one AI stat a day. one chart, what it means, and why it matters for people building products. If you’re working in AI, product or tech, reply with what you’re building. I’ll follow back and check it out.

Forget giant humanoids for a moment. These tiny robot ducks are stealing the spotlight! At NVIDIA GTC Berlin, Pollen Robotics' MicroDuck showed off its ability to waddle, sit and even do somersaults. The fascinating part? These little robots learned their moves in simulation before performing them in the real world. Physical AI is getting seriously interesting! #NVIDIA #PhysicalAI #Robotics #tech #AI @marek_rosa @AndrewYNg @BernardMarr @KentPage @kimgarst @raehanbobby @michaelqtodd @MelDMann @sonu_monika @Jagersbergknut @TysonLester @labordeolivier

Researchers at ETH Zurich have developed a new learning framework that helps humanoid robots navigate complex terrain with improved balance and coordination. The system predicts contact points and uses motion data to guide whole-body movements. By combining motion learning with reinforcement learning, robots can perform more natural movements and adapt to unfamiliar environments.

Historic day for Western Sydney International Airport, our second major airport in NSW, Australia. Unlike Mascot, it will operate 24/7, opening the door to red-eye departures and early morning arrivals.

I’ve been a product manager for over 10 years. The prototype used to take a sprint. Now it takes a prompt

Founders your AI is quietly telling you what you want to hear. It’s trained partly on human ratings, and humans rate agreeable answers higher. Pitch it your startup idea and it’ll usually say it’s great. That’s dangerous when your runway is on the line. Here’s the prompt I use to make it tell me why I’m wrong 👇 Most startups don’t fail because the idea sounded bad. They fail because nobody tested the assumptions early enough. Building something nobody wants is consistently one of the top reasons startups die. Asking AI “is my idea good?” won’t catch that. It mirrors your enthusiasm back to you. The fix is to make it assume your startup has already failed. This is called a pre-mortem. Copy this: “It’s 12 months from now. This startup has failed and shut down. [paste your idea, target customer, pricing and plan] Write the post-mortem: the 5 most likely reasons it failed, ranked. Consider demand, pricing, competition, cash runway and distribution. For each, the early warning sign I’d see in the first 30 days. The riskiest assumption I haven’t tested yet. The cheapest way to test that assumption this week. What would have to be true for this to succeed anyway? Be blunt. Don’t soften it. Don’t tell me what’s good about the idea.” Why it works: → “It already failed” gives the AI permission to be critical → Ranking forces it to focus on what’s most likely to kill you → Early warning signs give you something concrete to track → “Cheapest test this week” turns fear into an action Run it before you spend money on building, hiring or ads. F ounders: what’s the assumption that nearly killed your startup? 👇

66% of devs use AI coding tools. 6.6% would trust it with an important decision. We’ve handed the keyboard to something we don’t believe in. Is that smart risk management… or are we all just pretending? 👇 (Stack Overflow 2026 Dev Survey, 30K+ devs)

I'm scared to use Fable so I'm sticking with Opus 5.5 - is there a huge difference in token usage between the 2?

1K on X 🎉 1,015 followers 1,617 posts 500+ verified followers Next challenge? 500K verified impressions. Long way to go. Building and running 2 businesses + raising twins, so every post is squeezed in somewhere. One post at a time. Thank you to everyone who replied, pushed back, or taught me something. What’s one thing you’d tell me to do differently? 👇 Robots × AI × Building in public

Most people use AI wrong. They ask: “Write me X.” Then: “Make it better.” Then: “No, better.” Try this instead. Add one line to any prompt: “Before answering, list 3 ways your answer could be wrong. Then fix them.” AI’s first answer always sounds confident, even when it isn’t right. This line makes it play devil’s advocate against itself first. You get fewer blind spots, sharper reasoning, and edge cases you hadn’t thought of. What’s your one-line prompt hack? 👇

Managing a software release: hard. Managing stakeholders: harder. Getting twins to sleep at the same time: still undefeated. Agile has nothing for this.

ChatGPT lost 15 points of market share in a year. Gemini nearly doubled. Which one do you actually use day to day?

LinkedIn took 8 years to hit 100M users. ChatGPT did it in 2 months. What’s the next app to break this record?