Stephan Jaeckel

sam altman just dropped a blog post on why we're entering the era of the "idea guy" so, i wrote down 19 ways to come up with startup ideas in the AI age: 1. ask chatgpt: "give me 10 tedious workflows a [job title] does that ai could automate." rinse and niche down. 2. scroll your own chatgpt history. if you're repeating the same prompt weekly, that's a product hiding in plain sight. 3. search: site:reddit. com "is there a tool that" + your niche. you'll find users begging for tools that don't exist. 4. audit where you copy/paste between tools. that friction is a startup idea. 5. read ai tool youtube comments. people write what they wish the tool did, that's your roadmap. 6. join industry discords/slack groups. look for "does anyone have a tool for..." complaints. (can set up an automation for that on n8n) 7. use ideabrowser.com to find validated startup ideas from fb groups, subreddits, and google trends. use the ai agents to validate your own ideas (launch pricing rn @ideabrowser ) 8. study your calendar for recurring meetings with repetitive prep work. if you're copying templates each time, that's software. 9. browse fiverr for "i'll use chatgpt to..." services. those manual workflows are apps waiting to be built. 10. talk to local business owners about their daily excel updates. that's an ai agent waiting to be built. 11. browse g2 for top-rated legacy tools without ai. build the ai layer. 12. ask agency owners what clients beg for repeatedly. if it's annoying and repetitive, it's worth turning into software. 13. read comments on popular (70k+ views) youtube tutorial videos. look for friction points and niche applications for software ideas. 14. find multi-step tutorials using 8+ tools. build the one-click version. 15. scan linkedin job titles in growing industries. ask: "what would make this person 10x faster with ai?" 17. look for exploding job titles on linkedin. ask what workflows they're stuck doing manually. 17. find 2-star shopify/wordpress plugins with high usage. low stars + high usage = people care but experience sucks. or clone them with AI, focus on diff geographies and undercut on pricing. 18. join niche facebook groups and look for workflow hacking in comments. build the tool they're manually creating. 19. browse upwork for repeated ai + grunt work gigs. if multiple clients want it manual, it's ripe for productization. sam altman gets it. your biggest bottle neck is the right ideas (startup ideas/growth ideas). happy building. im rooting for you. we're all idea guys/gals now.

We’re past the “cool demo” phase. The real problem isn’t making robots look smart — it’s making them work in the real world, at scale. At @XSquareRobot, we made a very deliberate bet early on: stop stitching together brittle pipelines, and build a true end-to-end brain for the physical world. WALL-A isn’t a demo model — it’s a unified Vision-Language-Action foundation model that takes in raw perception and directly outputs action. No hand-designed modules, no shortcuts. That’s why the same model can run across different robots, different tasks, and messy environments — from hotels to logistics to real homes. Generalization isn’t a feature for us, it’s the starting point. What most people underestimate is this: embodied AI is not just a model problem, and it’s not just a hardware problem. It’s a closed-loop system — hardware, data, and model evolving together. That loop is the moat. EAIDC 2026 wasn’t just a conference. It was a signal that this field is shifting — from “can it work” to “how do we deploy it.” And increasingly, the ecosystem is building on top of what we’ve open-sourced. We’ve been on the end-to-end path long before it was obvious. We didn’t wait for validation — we built through it. Robots won’t stop at factories. They’ll move into services, homes — and eventually far beyond Earth. And when they do, the architecture we’re building now will be what scales with them.

We’re past the “cool demo” phase. The real problem isn’t making robots look smart — it’s making them work in the real world, at scale. At @XSquareRobot, we made a very deliberate bet early on: stop stitching together brittle pipelines, and build a true end-to-end brain for the physical world. WALL-A isn’t a demo model — it’s a unified Vision-Language-Action foundation model that takes in raw perception and directly outputs action. No hand-designed modules, no shortcuts. That’s why the same model can run across different robots, different tasks, and messy environments — from hotels to logistics to real homes. Generalization isn’t a feature for us, it’s the starting point. What most people underestimate is this: embodied AI is not just a model problem, and it’s not just a hardware problem. It’s a closed-loop system — hardware, data, and model evolving together. That loop is the moat. EAIDC 2026 wasn’t just a conference. It was a signal that this field is shifting — from “can it work” to “how do we deploy it.” And increasingly, the ecosystem is building on top of what we’ve open-sourced. We’ve been on the end-to-end path long before it was obvious. We didn’t wait for validation — we built through it. Robots won’t stop at factories. They’ll move into services, homes — and eventually far beyond Earth. And when they do, the architecture we’re building now will be what scales with them.

x.com/i/article/2039…

Israel’s success in developing a range of artificial-intelligence-enabled military technologies is underpinned by tie-ups with commercial entities, from Israeli start-ups to Palantir and big-tech corporations including Amazon, Google and Microsoft. These provide key enabling technologies, ranging from facial-recognition software and hardware that facilitate Israeli mass surveillance of Palestinians to commercial cloud servers used to host intercepted communications and train the Israel Defense Forces’ planned Arabic large language model. Explore the latest #ChartingMiddleEast analysis by Noor Hammad: go.iiss.org/47Bz0zy