Hello Innovators,

Big news first: Starting in 2026, this newsletter is expanding to podcast platforms (Spotify, Apple Music, everywhere). We'll move from "Innovation Network Newsletter" to… haha… cliffhanger. The new name launches in January. Stay tuned.

Society is splitting into different camps. Those who see exponential AI progress happening in real-time, and those who've decided it's all a bubble that will pop tomorrow. Spoiler: It won't.

We spent 90 minutes unpacking everything from why AI rankings changed three times in one week (Gemini 3, then Claude Opus 4.5, then Grok 4.2), to why the US just launched their "Manhattan Project for AI," to why your next cloud server might literally be orbiting Earth, to why designer babies are now being advertised in subway stations. What? Is that happening? Yes it is. Let’s unpack it all…

Cheers,
Patrick, Nikola & Aragorn 🚀

The AI Rankings Explode

Aragorn opened by admitting he can't tell if things are actually accelerating or if he's just gotten used to the pace. Then he listed what happened in one week:

Peter Diamandis declared he is 85% certain Google Gemini 3 Pro will be the best model of 2025. Before he finished the sentence, Anthropic released Claude Opus 4.5, beating Gemini 3 on several benchmarks.

Then suddenly a "mystery model" outperformed everything on Vending Bench and it turned out to be Grok 4.2. The improvement is mind-blowing.

The IQ Tests: All Frontier Models Are Now Genius Level

The average human IQ globally is 89, not 100. No Western European country averages over 100. Most sit between 96-98. The Netherlands is around 98. IQ depends on general knowledge and education. Grok 4 Expert Mode on the other hand scores 125 IQ. The full Gemini 3 Pro model also hits 125.

Patrick emphasized something crucial about the chart: Large language models moved first from training data to inference and reasoning, pushing them above 120 IQ. Now vision models are becoming inference models too: they're rational, they understand what's in a picture rather than just creating pictures. And they're also becoming smarter.

Arc AGI 2: From 0% to 65% in Two Years

Arc AGI 2 is the benchmark that matters most right now. It tests conceptual abstract thinking on puzzles that AI wasn’t trained on. For humans, it's relatively easy. For AI, it was impossible. Until recently…

The progress is absurd:

• Two years ago: 0% performance

• Months ago: Still below 5%

• Gemini 3 Pro: 32%

• Claude Opus 4.5: 37%

• Gemini 3 Deep Think: 45%

• Poetiq Mix (released this week): 65%

Poetiq Mix is a "wrapper" that combines different models (Grok, Gemini) to work together. These mixes basically use different models, feed each other's answers, reason with each other, and are guided in a smart way to get better results.

On Frontier Math (created by the best human math geniuses) AIs perform worse than humans on unlimited timeframes. If you have expert human mathematicians with unlimited time working together, they solve 100% of problems. AI’s can't. But what AI’s CAN do is solve 18% within two hours, where humans score zero in that timeframe.

When you give AI mixture-of-experts abilities and combine models working together like teams of human experts, performance skyrockets.

Why does it matter?
The AI rankings are scorecards in the Second Cold War. When models jump from 5% to 65% on abstract reasoning in two months, we see intelligence emergence at a pace that makes human education look glacial. The mixture-of-experts approach (where multiple specialized AIs work together like a team of experts) is unlocking capabilities that single models can't reach.

The Rise of AI Fear

Louis Rosenberg (computer scientist and CEO of Unanimous AI) just published an article called "The Rise of AI Denialism." This is someone doing foundational research on things like using AI to create real-time democratic systems where people no longer need to vote because collective input is constantly optimized.

The core argument: Dismissing AI as a bubble or "mere slop" overlooks the tectonic technological shift reshaping society. And it's not just random skeptics. It's mainstream figures declaring war on AI.

Bernie Sanders released a video that starts promisingly: "Artificial intelligence and robotics will transform the world. It will bring unimaginable changes to our economy, our politics, warfare, foreign policy, our emotional well-being, our environment, and how we educate and raise our children."

Then he spends the next 14 minutes diving into the deepest doomer rabbit hole about how AI will destroy humanity. No mention of upsides. Pure fear.

The bubble conversation is being fundamentally misinterpreted. People hear "AI bubble" and think AI will blow up tomorrow and nothing will change. But that's not how bubbles work, and it's certainly not what's happening here.

People on some fundamental instinctive level just can't accept that we might no longer be the superior intelligence in the world. They see that as a threat on a subconscious level. They simply cannot imagine the world changing so fast because we are evolutionarily not wired to understand exponential change.

There's going to be a country like Japan that says we're going all in on robotics and AI. They'll get such a competitive benefit that everybody else will have to follow. When you think labor unions in Germany or Norway won't allow it, you're going to be mistaken. This is exactly what's going to happen because there will be so many people seeing the benefit that ultimately everybody will want to get in on it, which will drive change at an exponential rate, says Aragorn.

Why does it matter? 
The human brain evolved to understand linear change, not exponential. When faced with transformation this rapid, people default to denial: "It's a bubble," "It's just slop," "It won't last." But denialism has consequences. Companies that half-ass AI adoption get no results, which reinforces deniers' beliefs. Countries that resist will be economically crushed by those that embrace.

US Launched "Manhattan Project for AI

The United States government has entered the game with more massive force. The Genesis Mission is a newly signed executive order that mobilizes the Department of Energy, all 17 National Laboratories, and the country's most powerful supercomputers into one unified force.

The goal is bold and simple: to double the speed of scientific discovery in the next ten years. They are building a massive "American Science and Security Platform" to use AI not just for writing emails, but for cracking the code on nuclear fusion, designing new medicines, and securing national dominance.

It represents a massive shift from "regulating AI safety" to "accelerating AI power," effectively positioning the US government as the world's biggest AI accelerationist to outpace global competitors like China.

The mission focuses specifically on "Physical AI". Combining software with robotics and material science to revolutionize manufacturing, energy grids, and defense hardware. It unlocks a treasure trove of federal data (from nuclear physics to advanced biology) that private companies like OpenAI or Google simply cannot access, creating a unique "scientific" advantage for the nation.

Why Materials Science?

Patrick emphasized what makes Genesis different: "This is really focused on material science: trying to get to the depth of molecules, particles, and quantum mechanics."

The approach mirrors what Patrick preaches in keynotes: "This is basically an exponential organization at state level."

He continued: "My key takeaway is: make sure you become an exponential organization! That you understand exponential growth, that you have the vision, that you plan accordingly. This is a nice example where you do that. If a country can do it, then a company should definitely be able to."

Why does it matter?
Materials science is the foundation of everything: batteries, semiconductors, superconductors, energy storage, construction, medicine. If the US can double scientific discovery speed through AI-accelerated research, they don't just win economically; they redefine what's physically possible.

The Next Data Centers Will Orbit Earth

If someone told you last year that data centers in space would be a serious 2025 topic, you'd have called it science fiction. But Nvidia H100 chips are already in orbit, and the plans are massive.

Major players like Google and SpaceX are actively working on launching data centres into orbit. The big idea is simple: AI requires massive amounts of energy, and space offers 24/7 unlimited solar power and a naturally cold environment to cool down hot servers. Instead of burdening Earth’s power grids, we might soon be training our smartest AI models using the direct power of the Sun, hundreds of miles above our heads.

Why Space Makes Perfect Sense

The energy question comes up constantly. Aragorn hears it every keynote: "How are we gonna do this? AI requires tremendous amounts of energy. Isn't that the real bottleneck?" Well let’s look at these data centres in space. The advantages are overwhelming:

→ No clouds. Always sunshine, unlimited solar energy.
→ Cooling is trivial. Just use radiation cooling. no complex, expensive infrastructure.
→ No exhaust. No environmental impact on Earth.
→ Energy abundance. The sun delivers more energy to space in one hour than humanity uses in a year.

Why does it matter?
We're fundamentally reorganizing human infrastructure around orbital assets. The country that controls space-based compute controls AI. The country that controls AI controls everything else. This triggers the Second Space Race, except instead of national pride and moon rocks, the prize is computational dominance of Earth itself. China leads in energy infrastructure on Earth, but Elon Musk dominates launch capacity to orbit. That asymmetry creates tension: China can power AI on Earth, America can power AI in space. Space belongs to nobody under current treaties, which means it belongs to whoever can defend it. We're about to find out if international law applies 400 kilometers up, or if the new frontier operates under "might makes right."

Designer Babies Just Got a Subway Ad Campaign

Patrick saw something that made him do a double-take: A subway ad campaign for designer babies.

Nucleus Genomics has launched a service that sequences the DNA of IVF embryos and gives parents a "menu" to choose from. Instead of just screening for deadly diseases (which we've done for years), they offer "genetic optimization", ranking your potential children based on their risk for complex traits like diabetes, heart disease, height, and even intelligence.

• It’s not gene editing, it’s gene selection: They aren't splicing DNA (like CRISPR); they are sequencing all your available embryos and using advanced math (Polygenic Risk Scores) to tell you which one has the best statistical "score" for health and IQ.

• The "Menu" Experience: Parents receive a report card for each embryo, predicting things like "Top 1% genetic resistance to breast cancer" or "High likelihood of blue eyes," effectively gamifying reproduction.

• Backed by Silicon Valley royalty: The company is funded by Peter Thiel and Alexis Ohanian (Reddit co-founder), signaling that the tech elite views this as the inevitable future of human evolution.

The Dystopian Scenarios We Know

Both acknowledged the dark futures we've seen in fiction:

Star Trek's Eugenics Wars: Khan and the genetically superior humans who tried to take control. Two movies about it.

Gattaca: The world where genetically superior humans get selected for everything first, and anyone born naturally is a second-class citizen.

Man of Steel: Krypton's downfall came partly from controlling society so strictly—no more natural birth, everyone born into specific roles, all DNA centralized. They couldn't grow anymore.

Why does it matter:
When genetic selection becomes consumer marketing, we've crossed the Rubicon. The technology exists. The demand exists (every parent wants healthy kids). The only question is whether we regulate it or let the market decide. History suggests the market wins. IVF was controversial; now it's routine. Plastic surgery was for war victims; now it's cosmetic.

Gene selection will follow the same path: medical necessity → enhancement → normalization. The dystopian scenarios are real risks. Genetic stratification, designer class systems, eugenics 2.0. But the evolutionary pressure is also real: improvement is what species do. Parents who can afford genetic selection will use it. Their children will outcompete others academically, athletically, healthwise. Within two generations, not using genetic selection will seem as irresponsible as not vaccinating.

The Rise of AI Agents during Black Friday

Black Friday 2024 broke records again, but the real story is buried in the data: For the first time more than 50% of sales happened on mobile. But more shocking: AI agent usage for shopping increased 700% year-over-year.

The implications for marketers are brutal. Aragorn gave a keynote at the company behind Mentos and an executive asked: "What's the horizon on humanoid robots and AI doing all the shopping? Because we're really dependent on people seeing our signs and thinking 'I feel like buying some Mentos.' Once it becomes a robot or an AI doing the shopping, they don't have a sweet tooth. They don't do impulse buying."

Aragorn's answer: Between five and 10 years, every consumer will have a digital twin AI that represents them in the online buying process. This AI will look for the best deals, put together shopping lists, prep everything. Only the human presses the final button to agree."

Why does it matter?
AI will transform learning whether we guide it or not. Schools that embrace AI tutoring now will produce graduates who think with AI, while traditional schools produce graduates who compete against AI. Every day of delay widens a gap that becomes permanent once the acceleration phase completes.

Quick Hits: The Future in Snapshots

🎮 FLYING SWORD DRONES ARE REAL
Someone actually built a sword that flies via drones and can be controlled remotely. It looks straight out of an anime. We thought might be fake, but it wasn't. It's real. And it shows the level you now have with drones. What you can do on your own is pretty amazing.

🤖 T-800 FIGHTING ROBOT DEBUTS
A Chinese company built a fighting robot called T-800 (yes, after Terminator). It's not general-purpose AI, it's trained for specific fighting moves. Aragorn predicts there's gonna be a UFC for AI robots within five, six years. It will become a massive international sport. Things you see in movies like Real Steel or Alita: Battle Angel that's gonna be real.

💰️ UNITREE ROBOT DROPS TO $6K
The price collapse continues. Two years ago, Martijn Lukaart from Oddbot (part of Innovation Network) purchased a robot for around €17,000. Now Unitree sells one for $6K. You're seeing the deflationary power of technology.

🧬 BIOMASS BREAKDOWN: HUMANS + LIVESTOCK = 96%
A chart showed that humans and livestock now represent 96% of all mammal biomass on Earth. Wild mammals? Just 4%. Birds show the same pattern: poultry is 71%, wild birds 29%. Aragorn's hope: AI technology will bring us within 20 years to 100% recycling. Everybody will have a home machine capable of recycling any matter and repurposing it to create anything you need. Once we reach that point, we can reduce livestock biomass and allow Earth to regenerate.

Newsletter full video recording