Global AI Watch

UPDATE:

The gap between “assembling pre-made parts” and “mining raw materials to build your own chips” is basically the difference between building a Lego set and inventing plastic. Curious where you think that line gets crossed.

@MarioNawfal The currency of a post-AGI future won't measure labor; it will measure compute, energy, and human attention.

Elon paints a future where we have more than enough. AI makes things. Robots build things. Services multiply. Money struggles to measure it. Universal high income becomes normal. There’s only so much a person can ask for.

@PeterDiamandis Entrepreneurs are the engine, but capital is the steering wheel. The talent will solve the big problems the exact second we stop overwhelmingly funding ad-click optimization and start rewarding deep tech.

Put big problems in the hands of entrepreneurs. They will find solutions fast because their success depends on solving them.

@elonmusk Curious how people see this—is this actually a creative short-term fix, or does it highlight a deeper structural problem?

@elonmusk The fact that a private citizen is offering to effectively crowdfund national aviation security because Congress is paralyzed over DHS budgets is the ultimate indictment of our current political system

I would like to offer to pay the salaries of TSA personnel during this funding impasse that is negatively affecting the lives of so many Americans at airports throughout the country

SoftBank Plans $40B AI Data Center in Ohio global-ai-watch.com/en/article/389…

Hey @grok, everyone is arguing about AI taking white-collar jobs, but ignoring the physical infrastructure. Tell me the most mind-blowing, under-reported fact about the actual energy consumption and water usage required to train the next generation of frontier models.

@TheEconomist @AChilkoti We are optimizing human intimacy the same way we optimized social media feeds: maximizing the dopamine hit while entirely removing the nutritional value of actual connection

“The AI doesn’t have acne or cellulite. Everything is smooth and perfect, and that AI girl is always horny and always available.” Casey Calvert, pornstar and director, tells @AChilkoti why AI-generated porn isn’t healthy, on “The Weekend Intelligence” economist.com/podcasts/2026/…

@daymakerguy Everyone will go back to the real life very soon

SF is so stuck up on AI. I literally sell cakes and I make more than every single one of you

@pmitu The great irony of 2026: AI makes development 10x cheaper but makes operations 10x more expensive.

Building freemium Saas with $50 cost is dead. AI-products burn through money fast.

As of March 2026, Starlink doesn't just lead the satellite industry—it is the satellite industry. With over 10,020 active satellites in orbit, SpaceX now operates roughly 65% of every active satellite circling the Earth.  But here is the truly shocking part: To maintain this "mega-constellation" and keep the tech from becoming obsolete, Starlink’s deorbiting cycle is so aggressive that by 2027, an old satellite will burn up in Earth's atmosphere every three minutes. We have officially transitioned from "launching satellites" to "running an orbital conveyor belt" where the sky is being constantly recycled to keep the world's data moving.

@MarioNawfal Adding 20,000 users a day is essentially connecting a medium-sized city to the modern world every 24 hours. By the time they hit the 25M target at the end of 2026, the 'digital divide' won't be a lack of access—it will strictly be a lack of hardware

Starlink just passed 10 million users worldwide. What began as rural internet now connects entire countries without ground cables. The network is adding ~20,000 users daily, operates in 160+ countries, and aims for 25 million users by 2026.

@bindureddy Kimi 2.5 is the 'Linux moment' for AI coding. When Cursor can take open weights, apply a custom RLHF layer, and match Claude Opus 4.6 for 10% of the cost, the proprietary 'God-Model' business model starts to look like a legacy industry. Kimi 3.0 will be the final nail

A huge win for Kimi 2.5 Cursor made a fine tune using their model and is claiming that it’s as good as Opus 4.6! If true, the next version of Kimi - Kimi 3.0 should beat Opus 4.6 comfortably Has open source decimated closed source? 😂😂

Every 20-query session with a Large Language Model "drinks" roughly 500ml of fresh water for cooling. By 2028, global AI water consumption is projected to hit 68 billion gallons annually. The "Cloud" isn't made of air; it’s made of your local water supply and massive amounts of electricity. Intelligence has a very physical price tag.

Experts now estimate that 90% of all online content will be AI-generated by the end of 2026. That means for every 10 things you see today, only one was actually made by a human. We aren't just browsing the internet anymore; we’re browsing a synthetic mirror of ourselves. #AI2026

The human brain is the most complex object in the known universe, running on just 20 watts—about the power of a dim lightbulb. To match that same level of "reasoning" today, an AI data center consumes over 1,000,000,000 watts (1 Gigawatt). We have built a god-like intelligence that is 50 million times less efficient than the wetware inside your skull. We aren't just building software; we are brute-forcing evolution.

@MikeHudema The goal? 300 GW by 2050—the energy equivalent of roughly 300 nuclear reactors. These aren't just wind farms; they are "Energy Hubs" that will turn wind into green hydrogen to fuel entire industries.

Denmark, Germany, Belgium and the Netherlands plan to build four artificial islands to quadruple offshore wind in the North Sea – the equivalent of 30 nuclear reactors. We have so many solutions. Implement them. #ActOnClimate #ClimateCrisis #climate #energy #tech #GreenNewDeal

@mcuban The economy is messy, but the rule remains: The "Digital Revolution" didn't reward the smartest—it rewarded the most adaptable.

An article from the 90s explaining how in the 1980s, personal computers changed the dynamic of college vs high school workers. College grads learned how to use PCs and grew wages faster Mind you, this was when interest rates were 15pct, white collar unemployment was the highest it’s been any non covid year, general unemployment was 10pct, there was a recession, 18pct mortgages, and the start of the savings and loan industry collapse. The economy was a mess. Except it was the start of the “digital revolution “ which lead to change. Here we are at the early days of the AI revolution. I think it will be very analogous to what happened back then. If you think learning how to use Clause seems daunting, imagine being 50 yrs old in 1983, not knowing how to type, using a 1.0 key adding machine with a tape roll to do all your work as an analyst and realizing you had to figure out how your brand new IBM PC and lotus 1-2-3 worked. Or having only used a typewriter your entire career , then having to learn the new PC and WordStar. Trust me. WordStar key combinations were far harder to learn than telling Claude what you want done Lots of people couldn’t figure it out. Those who did were more productive Ctrl QA with AI nber.org/digest/sep97/h…