Genesis Mission
So "Manhattan Project" for AI is already in place like I said
Hey there. Wanted to share recent news and updates.
On Monday, Trump signed an executive order that most people scrolled past.
No flashy product launch. No viral demo. Just a government document with a name that sounds like a sci-fi movie.
Genesis Mission.
It’s the largest mobilization of federal scientific resources since the Apollo program.
And possibly the most critical AI decision any government has made.
While everyone was arguing about which chatbot writes better code, the United States quietly decided to treat artificial intelligence like nuclear weapons.
I wrote about AI companies wielding nuclear-level power before
Training costs in the hundreds of millions. Weekly release cycles that feel like an arms race. Engineers who don’t sleep on weekends.
“The power of the sun, in the palm of my hand.”
That was a metaphor.
Now it’s literal.
The same national laboratories that built the atomic bomb are entering the AI race. Oak Ridge. Los Álamos. Lawrence Livermore. The places where America learned to split atoms are now learning to train models.
This is that story.
What “Genesis” Actually Is
Here’s the simple version:
America is unifying all its scientific firepower under one AI-powered roof.
The Department of Energy operates 17 national laboratories. These aren’t ordinary research centers. These are the institutions that:
Built the atomic bomb (Los Alamos, Oak Ridge)
Cracked the human genome (Lawrence Berkeley)
Run the most powerful supercomputers on Earth (Oak Ridge’s Frontier, Argonne’s Aurora)
Conduct classified research that shapes national security
Until now, they worked separately. Different systems. Different datasets. Different priorities. Different bureaucracies.
Genesis changes everything.
All 17 labs are merging their supercomputers, scientific data, and AI models into a single integrated platform.
Michael Kratsios, the White House science advisor, called it:
“The largest marshaling of federal scientific resources since the Apollo program.”
And:
“The largest collection of scientific data in the world.”
That’s not marketing. America has been collecting scientific data for 80 years. The Manhattan Project. The Human Genome Project. Decades of climate research, materials science, nuclear physics, and biomedical studies.
All of it. Unified. AI-powered.
The Audacious Promise
Genesis isn’t about making scientists 10% more productive.
It’s about compressing discovery timelines from years to days. Maybe hours.
Here’s what the platform will do:
Automate Experiment Design
Instead of scientists spending months designing studies, AI generates optimal experimental protocols. It identifies variables humans might miss. It suggests approaches no one considered.
Run Simulations at Unprecedented Scale
Problems that would take years to model, protein folding, fusion plasma dynamics, and materials behavior under extreme conditions, become solvable in days.
Generate Predictive Models
AI doesn’t just analyze past data. It predicts outcomes. Which drug compounds will work? Which materials will fail? Which nuclear configurations are stable?
Power Autonomous Laboratories
Robotic systems that can test hypotheses, analyze results, and iterate. While you sleep.
One White House official put it bluntly:
“We’re going to understand the exact chemistry and biology of diseases that are death sentences today. We’ve seen progress the last five years. It’s nothing compared to what we’re going to see in the next five years.”
That’s the ambition. Science at machine speed.
The Strategic Priorities
The priority areas tell you exactly what America considers existential:
🧬 Biotechnology Drug discovery. Disease prediction. Synthetic biology. Pandemic preparedness. COVID made clear that biotech is national security.
⚛️ Nuclear Energy: Both fission and fusion. America needs energy independence. Genesis will accelerate reactor design, safety modeling, and the long-shot quest for commercial fusion.
🚀 Space Exploration: Someone has to compete with China’s moon ambitions. AI-powered mission planning, materials science for spacecraft, and autonomous systems for deep space.
💻 Quantum Computing: The next computational paradigm. Whoever cracks practical quantum computing first gains advantages in cryptography, optimization, and simulation that could last decades.
🔬 Semiconductors: The chokepoint of the entire AI race. Every model, every training run, every inference run on chips. Controlling chip design and manufacturing is controlling the future.
🛡️ National Security Weapons design. Cryptography. Cyber defense. Intelligence analysis. The applications they’re not talking about publicly.
This isn’t a research grant program.
This is a country deciding that AI-powered science is a matter of national survival.
In “AI -> AGI -> ASI,” I wrote about the transition from narrow AI to artificial general intelligence. The security implications. The arms race dynamics. The existential stakes.
“With great power comes great responsibility. The advent of AGI raises significant ethical, security, and socio-economic challenges.” I said.
Genesis is the government’s answer to those challenges. They’re not waiting for AGI to arrive. They’re building the infrastructure to control it.
The People
Let’s talk about who’s actually in charge.
Michael Kratsios: The Operator
Age: 38
Background: Started at Peter Thiel’s venture fund. Became America’s first Chief Technology Officer under Trump’s first term.
Then something unusual happened: a tech guy moved to the Pentagon.
Kratsios became Deputy Secretary of Defense for Research and Engineering. That’s the third-highest position in the entire Department of Defense.
Under his command:
DARPA (the agency that invented the internet)
Missile Defense Agency
The entire military laboratory system
A budget of $106 billion
He’s not a bureaucrat. He’s commanded more R&D resources than most countries have in total.
David Sacks: The Connector
Background: PayPal Mafia. Built Yammer, sold it to Microsoft for $1.2 billion. Currently runs Craft Ventures with investments in SpaceX and Airbnb.
Close friend of Elon Musk.
Interesting trajectory: In 2021, Sacks publicly said Trump “disqualified himself” after January 6th. By 2024, he’d raised $12 mil. for Trump’s campaign.
Now he’s the “AI and Crypto Czar”, a role that doesn’t require Senate confirmation. His job: coordinate the private sector around Genesis.
These aren’t career government employees waiting for retirement.
These are operators who’ve built companies and commanded massive resources. They know how to move fast.
The question is whether they can also prompt the government to move quickly.
The Private Sector Alliance
Genesis launched with immediate private sector commitments:
Nvidia: The GPU monopoly powering all AI training
Dell: Enterprise infrastructure
Oracle: Cloud and database systems
Anthropic: Leading enterprise AI, makers of Claude
The model is elegant:
Companies fund data center construction on government land
National labs provide computing power, scientific data, and research talent
Everyone shares access to the resulting AI infrastructure
It’s the same playbook that built the internet. The government provides the foundation. The private sector builds on top. Everyone benefits.
But here’s what makes Genesis different from Stargate and other recent AI mega-announcements:
This isn’t about building new stuff. It’s about coordinating what already exists.
America already has:
✅ The world’s best supercomputers
✅ The most comprehensive scientific datasets
✅ The leading research institutions
✅ The top AI companies
The problem was fragmentation. Siloed data. Incompatible systems. Bureaucratic barriers.
Genesis is the integration layer.
That’s harder than it sounds. And more valuable than most people realize.
What’s Overlooked
Problem #1: No Concrete Funding
White House officials said they’ll “continue to invest increasing amounts for the success of the mission” with help from Congress.
Translation: They want money but haven’t secured it.
Coordination is free. Building is not.
The Department of Energy already operates world-class facilities. But scaling them? Integrating them? Building the software layer that makes it all work? That costs billions.
If Congress doesn’t fund Genesis properly, it becomes another government initiative with great slides and no outcomes.
Problem #2: Data Access Is Hard
Genesis promises to give private companies access to government scientific data.
Sounds simple. It’s not.
Government data is:
Messy – Collected over decades with inconsistent formats
Classified – Much of it has national security restrictions
Regulated – Privacy laws, export controls, agency-specific rules
Scattered – Across agencies with different systems and incentives
The legal and bureaucratic work required to actually share this data could take years.
Kratsios knows this. He spent four years navigating Pentagon bureaucracy. The question is whether Genesis can move at startup speed inside government structures designed for caution.
Problem #3: Talent Competition
The same AI researchers Genesis needs are being recruited by Google, Anthropic, OpenAI, and every other tech company with billions to spend.
Government salaries can’t compete. Government pace can’t compete. Government bureaucracy can’t compete.
Genesis will need creative solutions: joint appointments, contractor relationships, or simply relying on the national labs’ existing workforce.
None of these problems are fatal. But they’re real.
And they’ll determine whether Genesis delivers on its promise or becomes an expensive coordination exercise.
The China Factor
Let’s be direct about why Genesis exists.
China.
Beijing has its own version: the National Integrated Computing Network, coordinated by the Peng Cheng Laboratory. Same concept. Unify national computing resources. Apply AI to scientific research. Accelerate discovery.
But China has two massive disadvantages:
Disadvantage #1: The Compute Gap
The numbers are stark, Global AI Computing Capacity:
United States is ~75%
China is ~15%
Export restrictions on Nvidia chips are already hurting Chinese developers. They can’t buy the best hardware.
They’re forced to use inferior alternatives.
Huawei’s Ascend 910C is catching up. But it’s still behind. The gap matters more as models become larger.
Disadvantage #2: The Data Gap
America’s national labs have been collecting scientific data for 80 years.
Nuclear physics from the Manhattan Project era
Genomic data from the Human Genome Project
Climate data from decades of environmental monitoring
Materials science from defense research programs
This institutional knowledge is irreplaceable. China can’t download it. They can’t replicate it. They can only try to generate their own, which takes time they may not have.
Genesis plays directly to these American advantages.
More compute. Better data. Faster integration.
If it works, America extends its AI lead in exactly the domains that matter:
Biotech (pandemic response, drug discovery)
Nuclear (energy independence, weapons modernization)
Quantum (cryptographic advantage)
Defense (autonomous systems, intelligence analysis)
If China responds, and they will, expect massive investment in domestic chip production and accelerated lab integration.
The AI race just became a science race.
Why again
Genesis didn’t happen in a vacuum.
The same week Trump signed the executive order, the AI industry experienced its most consequential seven days in years.
🔵 Google Released Gemini 3
First model to break 1500 Elo on LMArena. The highest score any AI model has ever achieved. I am really enjoying using and building a product. I will let you know in the upcoming editions.
State-of-the-art in:
Reasoning (91.9% on GPQA Diamond)
Mathematics (95% on AIME 2025 without tools, 100% with code execution)
Multimodal understanding (81% on MMMU-Pro)
Coding and web development
The company that everyone wrote off two years ago is now leading.
🟠 Anthropic Dropped Claude Opus 4.5
Best coding model in the world. 80.9% on SWE-bench Verified. The first model to break 80%.
Anthropic now owns 32% of the enterprise AI market. OpenAI dropped to 25%, half of what they held in 2023.
The safety-focused underdog became the enterprise leader.
💰 Microsoft and Nvidia Invested in Anthropic
$350 billion deal.
Nearly doubled Anthropic’s valuation overnight.
The biggest AI investment partnership ever announced.
📈 The Market Shifted
According to Menlo Ventures, enterprise LLM spending doubled from $3.5 billion to $8.4 billion in just six months.
AI isn’t a future bet anymore. It’s current infrastructure.
This Is The Context For Genesis
Two years ago, AI was a tech industry story. Interesting. Promising. Contained.
Now it’s a national security story.
The companies building the best AI models will shape:
Scientific discovery for decades
Economic productivity globally
Military capability for every nation
Governments can’t watch from the sidelines anymore.
Genesis is America’s answer to that reality.
It’s not a research program. It’s not an innovation initiative. It’s a strategic mobilization.
The same urgency that drove the Manhattan Project. The same scale that drove Apollo.
Applied to artificial intelligence.
The Babylon Update
Quick aside for longtime readers.
In “Babylon Fall,” I wrote about Google’s decline. How the company that defined the internet era was stumbling in AI. How Bard was embarrassingly bad. How the empire was crumbling while OpenAI ate their lunch.
I was wrong.
Gemini 3 doesn’t just compete. It leads.
The model completes 10 to 15 coherent reasoning steps where previous models stumbled at step 5 or 6. That’s not incremental improvement. That’s a different category of capability.
And Google’s real advantage isn’t the model. Its distribution.
Gemini 3 launched simultaneously across:
Google Search AI Mode (2 billion users)
Gemini App (650 million users)
Google Cloud services
Day one. Everywhere.
OpenAI has 800 million weekly users. Impressive. But Google has 2 billion people accessing Gemini through Search alone.
Gary Marcus, the AI skeptic, wrote it plainly:
“OpenAI has basically squandered the technical lead it once had. Google has caught up.”
The tables turned faster than anyone predicted.
Google’s back. OpenAI is fighting for relevance. Anthropic’s dominating enterprise.
And now the US government is entering the game with the biggest mobilization of scientific resources in half a century.
OpenAI
While Google was releasing frontier models and the government was launching its Manhattan Project for AI, OpenAI released a shopping assistant.
Yes, really.
ChatGPT now has a dedicated “Shopping Research” feature. You ask it to find “the quietest cordless stick vacuum for a small apartment,” and it builds you a personalized buyer’s guide.
The feature:
Asks clarifying questions about the budget and preferences
Searches across retailers (except Amazon, which blocked their crawlers)
Compares specs, prices, and reviews
Generates a structured recommendation
It’s available to all users, including free accounts. Nearly unlimited usage through the holidays.
My take
OpenAI is losing the enterprise race to Anthropic. They’re losing the benchmark race to Google. Their market share dropped from 50% to 25% over a two-year period.
So they’re pivoting to consumer monetization.
Shopping is a $5 trillion market. If ChatGPT can capture even 1% of purchase decisions, that’s real revenue. No enterprise sales team required.
It’s innovative business. But it’s also a signal.
The company that was supposed to build AGI is now helping you find vacuum cleaners.
In “Replacing Governments,” I wrote about tech companies becoming powerful enough to rival nation-states. I wondered if AI labs would reshape society itself.
I didn’t expect one of them to become a shopping concierge first.
The Productivity Revolution Is Already Here
Speaking of Anthropic.
While everyone debates future AI capabilities, Anthropic just published research that quantifies what’s already happening.
They analyzed 100,000 real Claude conversations to measure actual productivity gains. Not benchmarks. Not demos. Real work.
The headline numbers:
AI reduces task completion time by 80% on average
Tasks that would take 90 minutes without AI take about 18 minutes with it
The median task would cost $55 in human labor to complete manually
But here’s what most people will miss: the variation across occupations is massive.
Time Savings by Occupation
The Macro Implications
Anthropic extrapolated these findings to the entire US economy.
Current AI models could increase US labor productivity by 1.8% annually over the next decade.
That would double the productivity growth rate we’ve seen since 2019.
To put this in perspective:
US productivity growth averaged 2.1% annually since 1947
It’s been 1.8% since 2019
Adding another 1.8% from AI would bring us to 3.6% total
That’s not incremental improvement. That’s a structural shift in economic growth.
The research reveals something crucial that most analysis misses:
As AI accelerates some tasks, others become bottlenecks.
A software developer might complete coding tasks 80% faster, but still spends the same time in meetings, coordinating installations, and supervising other engineers. Those un-accelerated tasks now represent a larger share of the job.
A teacher might prepare lessons 85% faster, but still spends the same time enforcing classroom rules and sponsoring extracurricular clubs.
The tasks AI can’t help with become a constraint on growth.
This is the real insight. Productivity gains aren’t uniform. They create new bottlenecks. The most valuable workers will be those who excel at the tasks AI can’t accelerate.
The Contribution Distribution
Which occupations drive the most aggregate productivity gain?
Software Developers – 19% of total gain
General & Operations Managers – 6%
Market Research Analysts – 5%
Customer Service Representatives – 4%
Secondary School Teachers – 3%
Meanwhile, restaurants, healthcare delivery, construction, and retail contribute almost nothing to aggregate productivity gains. Not because AI can’t help, but because those jobs have so few tasks that current AI can accelerate.
Anthropic explicitly states:
This analysis assumes AI capabilities stay the same for 10 years.
They won’t.
The models are improving rapidly. The 1.8% productivity boost is a floor, not a ceiling. And it doesn’t account for how companies will restructure around AI capabilities.
Historically, the biggest productivity gains from new technologies came not from speeding up old tasks, but from fundamentally reorganizing how work is done.
We’re still in the “speeding up old tasks” phase.
The reorganization phase hasn’t even started.
What This Means For You
This is a long read, so here is TLDR.
If you’re building, investing, or just paying attention, five things matter:
1. The Models Are Converging
Gemini 3, Claude Opus 4.5, and GPT-5.1 are all remarkably capable. The gap between “best” and “second best” is shrinking every month.
Implication: Select models based on your specific use case, rather than relying on benchmark rankings. The “best” model changes quarterly. Your integration architecture shouldn’t.
2. Enterprise Adoption Is Exploding
The LLM market doubled in size from $3.5 billion to $8.4 billion in just six months. That’s not hype. That’s companies deploying AI in production.
Implication: If you’re not integrating AI into your workflows, your competitors are. This isn’t about experimentation anymore. It’s about implementation.
3. Government Is Now A Player
Genesis means federal resources flowing into AI infrastructure at an unprecedented scale. National labs, supercomputers, and scientific data are being unified and made accessible.
Implication: If you’re in biotech, energy, materials science, defense, or any adjacent industry, pay attention. Contracts, partnerships, and opportunities are coming.
4. The Consumer AI Battle Is Just Starting
Google has distribution. OpenAI has brand recognition. Anthropic has enterprise trust.
Implication: The winner of the consumer AI race hasn’t been decided yet. If you’re building consumer products, the platform you choose matters more than ever.
5. Your Value Is In The Bottlenecks
AI is accelerating 80% of knowledge work. The remaining 20%, judgment, coordination, relationships, and physical presence, become more valuable by contrast.
Implication: Identify which parts of your work AI can’t accelerate. Double down on those. They’re your competitive moat.
Post-Credit Scene
The AI race used to be about which startup had the best demo.
Now it’s about which country can coordinate trillion-dollar resources most effectively.
That’s a different game entirely. Genesis is America’s bet that government-scale coordination, combined with private sector speed, can accelerate scientific discovery faster than any single company or country could alone.
It might work. It might get buried in bureaucracy.
But the ambition is real. The resources exist. And the people running it have actually operated at scale before. If Genesis delivers even a fraction of its promise, the implications ripple everywhere:
Drug discovery accelerates
Energy technology advances
Materials science breakthroughs compound
Defense capabilities shift
We’re watching a country decide that AI isn’t just a technology.
It’s infrastructure. It’s a strategy. It’s survival.
And they’re treating it accordingly.
Ok you made it here, so here is your post-credit things worth your time this week:
📺 “Pluribus” (Apple TV+) – Must watch.
Vince Gilligan’s new series starring Rhea Seehorn just broke Apple TV’s viewership records. The premise: humanity suddenly joins a hive mind of forced happiness, and Carol is one of the few immune. She has to save the world from bliss. It’s part sci-fi, part black comedy, part psychological thriller. From the creator of Breaking Bad and Better Call Saul, 100% on Rotten Tomatoes. The themes of individuality vs. collective consciousness hit different when you’re reading about governments coordinating AI at a civilization scale.
📖 “The Maniac” by Benjamín Labatut A novel about John von Neumann and the birth of computing. Same labs. Same ambitions. Different century. Eerily relevant given Genesis. If you want to understand the human drama behind scientific mobilization, start here.
🎧 “Acquired” Podcast: Nvidia Episode The definitive breakdown of why chips matter so much to the AI race. Explains exactly why China’s compute disadvantage is so significant, and why Nvidia’s position is so powerful.
📺 “Oppenheimer”
Yes, again. Watch it, thinking about Genesis. Same national laboratories. Same tension between scientific ambition and national security. Same question: what happens when you concentrate this much power in the service of a single goal?
📄 Anthropic’s The full study on AI productivity gains. Shows exactly how 100,000 real conversations translate into 80% time savings and potential doubling of US productivity growth.
📄 Menlo Ventures 2025 LLM Market Report: The actual data on enterprise AI adoption. Shows exactly how Anthropic overtook OpenAI, why Google’s gaining fast, and where the $8.4 billion is flowing.
Thanks for reading.
Vlad