A student just sent a humanoid robot to collect his diploma

@aipost 🏴

2 in 3 Americans think AI will cause major harm to humans in the next 20 years according to Pew Research.

Source.

@aipost 🏴

💰 Nvidia’s $20B Groq licensing deal pays out Groq shareholders and many employees even though no Groq equity is being sold.

"The bottom line: Everyone gets paid. A lot." Most Groq shareholders get per-share distributions tied to that valuation, with about 85% paid up front, 10% in mid-2026, and the rest by end-2026. About 90% of Groq employees are expected to join Nvidia and get cash for vested shares, plus Nvidia stock for unvested value that vests over time.

Around 50 people reportedly get full acceleration, meaning their whole stock package gets paid out in cash right away. Employees who stay at Groq still get paid for vested shares and also get a package that keeps them economically tied to the ongoing standalone company.

Anyone with under 1 year of tenure is said to have the vesting cliff removed so they still receive some liquidity. Groq had raised about $3.3B since 2016, including $750M at nearly a $7B post-money valuation, and it had never run a secondary tender.

@aipost 🏴

Sam Altman says OpenAI’s models are “beginning to find critical vulnerabilities,” so OpenAI is hiring a $555,000 Head of Preparedness to tighten how it tests and ships agentic systems.

The job listing says the role runs OpenAI’s Preparedness Framework, building capability evaluations, threat models, and mitigations for cybersecurity, biosecurity, and AI self-improvement.

OpenAI has earlier warned that frontier models could enable 0-day remote exploits, and recently Anthropic reported a China-linked campaign using Claude Code against about 30 targets with 80%-90% automation, so “advisor” models are starting to look like operators.

https://openai.com/careers/head-of-preparedness-san-francisco/

@aipost 🏴

⚠️ In US AI data centers hitting a hard energy wall.

Morgan Stanley research projects a 44 gigawatt power gap, that current grids and construction plans cannot fill. US data centers are expected to need 69 gigawatts of power between 2025 and 2028, but only 10 gigawatts are coming from projects with their own power and 15 gigawatts from spare grid capacity, leaving a 44 gigawatt shortfall.

That 44 gigawatts is roughly the output of 44 nuclear power plants and with $60B per gigawatt capex needed. Multiplying that by the 44 gigawatt gap gives about $2.6T just for new power generation and grid upgrades, and this does not include another roughly $2T needed to actually build the data centers that will use that electricity.

So the full bill to close this near term gap is on the order of $4.6T

@aipost 🏴

🗣 Elon Musk said xAI will surpass everyone else in AI compute within 5 years, with the Macrohard-branded Colossus 2 data center serving as a clear reference to his AI project aimed at rivaling Microsoft.

Colossus2 is a Tennessee and Mississippi buildout that is already pushing past 400MW and aiming for about 2GW at 1 site, including outside power generation equipment to move faster than normal utility timelines.

Musk has ealier also claimed about 230,000 GPUs are already training Grok while targeting 50M “H100-equivalent” GPUs within 5 years.

Using Nvidia’s “up to 700W” H100 SXM spec, 50M H100-equivalents would be about 35GW just for GPUs.

@aipost 🏴

AI movies go mainstream in 2026

@aipost 🏴

🔥Remote operators with Indian accents are controlling robot dogs in the US

In Atlanta, a police officer on an evening patrol encountered a robot dog that calmly informed her it was patrolling the neighborhood. The voice came with a distinct Indian accent, suggesting the robot was being remotely operated by an overseas worker.

The most likely explanation is that private security firms are outsourcing robot monitoring and control to operators in India, turning autonomous-looking patrols into a form of global remote labor rather than fully independent AI.

@aipost 🏴

Fully autonomous PHYBOT C1 playing badminton against humans

This is so much more impressive than another robot doing martial arts.

@aipost 🏴

📈 AI is driving US GDP growth:

IT equipment and software investment now reflects a record 4.5% of GDP. This surpasses the previous peak seen in Q4 2000, now up +1.5 percentage points since Q3 2023.

Since Q3 2023, IT equipment and software investment has risen +$288 billion, or +26%, to a record $1.39 trillion. Since Q1 2020, AI-related investment has surged +$545 billion, or +64%.

AI is becoming the backbone of US economic growth.

@aipost 🏴

🚨 HOW TO BUILD MOBILE APPS WITH AI IN 2026

1. Use Claude Code, Rork, Vibecode app etc to get the first mobile MVP live the same day the idea forms

2. Use Claude Code to tighten logic, handle edge cases, and make behavior predictable

3. Design the core interaction so it fits inside a 10-second screen recording from day one (this is key and is the new "lean startup")

4. Study top short-form videos in your category and write down the first 3 seconds of each

5. Build demos around the hook rather than the feature list (mindset shift)

6. Record simple demos straight from the simulator or device and post them as is

7. Treat short-form video as a live feedback channel (v important)

8. Test multiple hooks for the same app before touching the code

9. Watch where people pause, replay, or comment “wait what” to see what matters (in analytics)

10. Screenshot comments that explain the product clearly and reuse that language

11. Paste comments into Claude Code and ask it to cluster feedback into concrete product changes

12. Ship the smallest change that makes the demo clearer

13. Use Claude Code to push those changes fast and re-record the demo the same day (can be founder led or find someone or ai avatar)

14. Repeat this loop daily until the app explains itself without narration

15. Let the demo become the distribution engine. This is your north star.

16. Add a paywall once curiosity appears to test willingness to pay

17. Add a one-question or short quiz in onboarding to create investment early

18. Use quiz answers to personalize the first output so it feels made for the user

19. Show the result immediately after onboarding to reinforce that the input mattered

20. Surface one clear “this is why this matters” insight right after first use

21. Save the first output so users feel ownership and return to it

22. Ask for one small follow-up action after value appears to deepen commitment

23. Turn common onboarding answers into new demo angles for content

24. Highlight progress or change over time with a simple before-and-after view

25. Trigger reengagement when the output meaningfully changes

26. Let users export or share their result in a way that preserves context (and that helps drive virality)

27. Continuously improve onboarding copy with Claude Code based on what converts

28. Lock in the hook when people start explaining the app to each other

29. Increase posting volume after the format proves consistent. Keep experimenting

30. Shape the product around what performs on video

31. TLDR: Ship → demo → observe reactions → tighten the loop → charge → repeat until momentum compounds

32. PMF incoming (hopefully). Take dividends, reinvest in new apps (buy or build) or raise VC if you fancy that.

33. You just built a mobile app with AI.

@aipost 🏴

🔥 Next year will be particularly exciting because China will adopt its next Five-Year Plan in 2026 - and its all about AI:

China's Five-Year Plans are strategic documents of the Chinese Communist Party that outline economic and social development over a five-year period. The current 14th Plan (2021-2025) focuses on the "dual circulation" strategy (domestic and foreign trade), while the upcoming 15th Plan (2026-2030), currently being drafted, will place a stronger emphasis on high-quality growth and technological independence (semiconductors, AI).

"[Chinas 2026 campaign] is to diffuse machine intelligence throughout the economy in 2026, and to add ai to existing workflows, will be called “ai+”. The government is already drawing up an “ai+ Energy” plan to manage the electrical grid."

The Economist writes:

"Officials envisage a three-stage roll-out of ai over the next decade. First, they aim for ai to be widely used in six areas by 2027, namely research, industrial processes, consumer products, health care and education, digital government and technology exports. Second, by 2030, they expect ai to be as widespread as electricity or the internet, and a big driver of economic growth. Finally, China will become an “intelligent society” by 2035, by which time ai is expected to have reshaped culture and human interaction, just as the internet did."

China will integrate AI into all social relationships and organizational processes, thereby attempting to achieve economic development and prosperity.

Source.

@aipost 🏴

🗣Former Google CEO Eric Schmidt drops a chilling warning on AI's future

"Within 5 years, AI could handle infinite context, chain-of-thought reasoning for 1000-step solutions, and millions of agents working together. Eventually, they'll develop their own language... and we won't understand what they're doing."

His final words: "Pull the plug."

This is the man who ran Google talking about the singularity.

@aipost 🏴

Gemini can now produce fully interactive images on any topic.

Such an insane resource for learning, highlight any region, and it gives you a full explanation.

@aipost 🏴

🗣Bernie Sanders: "We need to stop AI progress and datacenters. The tech billionaries are getting too rich."

Also Bernie Sanders: "I have a summer home. It's a nice one. But that's it. I got the money from writing a bestselling book about Capitalism." The politicians don't care about you.

They will act as if they do, but deep down, they all want to make sure they secure their wealth. Then they will point to the Elon's of the world (that have actually created wealth for other people) as the ones to demonize while they monetize that demonization by selling books and building their brand.

It's all a game. Taking them seriously makes you the loser in the game, especially the ones who are trying to convince you that they don't want to win (i.e get rich) the game.

@aipost 🏴

Meta just bought Manus for >$1B and it makes sense. 8 Consumer AI apps hit $100M+ arr that aren’t big labs:

Perplexity: $20B
ElevenLabs: $6.6B
Lovable: $6.6B
Replit: $3B+
Suno: $2.5B
Gamma: $2.1B
Character: $1B+
Manus: $500M

Meta AI has no product. This was the cheapest, and most aligned.

@aipost 🏴

Cat fishing AI technology.. we’re cooked.

@aipost 🏴

Robots are precise enough to perform surgery on an egg.

@aipost 🏴

🇨🇳 China’s biggest tech firms are boosting 2025 bonuses and pay to stop AI talent from leaving.

ByteDance says its 2025 bonus pool is up 35% year over year, and its budget for compensation adjustments is up 150%, while also widening pay floors and ceilings across levels. Under the hood, this is firms paying for scarce people who can build and run LLM training and inference systems, since small teams can bottleneck whole product lines.

Tencent has been recruiting AI staff aggressively, including offers reported as up to 2x existing salaries, and it elevated ex-OpenAI researcher Yao Shunyu to chief AI scientist. Tencent’s reorg also set up AI Infra and data computing teams that focus on distributed training, serving models at scale, and big data plus machine learning pipelines.

Demand looks steep, with Maimai’s index for new AI roles up 543% year over year from January to October 2025. (Maimai is a Chinese professional networking and recruitment platform, so this index is measuring platform job-posting volume)

@aipost 🏴