Chonsy

Obsessed with AI | Researcher & Developer

[Chonsy @ChonsyAI]

An AI just got access to the US biodefense grid. It will detect the next pandemic before any government does

OpenAI briefed the White House before the public knew. All integrated. Rosalind now runs inside US pandemic defense. It knows before you do

Here is what actually happened

On May 29 OpenAI launched Rosalind Biodefense. A program that gives GPT-Rosalind - their frontier AI model built specifically for life sciences - controlled access to US government biodefense infrastructure

Not a research partnership. Not a pilot. Trusted access to the institutions that protect humanity from the next pandemic

Lawrence Livermore National Laboratory is using it to design and evaluate medical countermeasures for emerging biological threats. Johns Hopkins Applied Physics Laboratory is integrating it into a protein engineering platform to screen therapeutic candidates. The Coalition for Epidemic Preparedness Innovations is applying it to accelerate vaccine development against epidemic threats including the current Ebola outbreak

Here is what GPT-Rosalind actually does

Epidemiological modeling. Early detection support, Outbreak response planning. Diagnostics. Vaccine development acceleration. Literature synthesis. Protocol design. It processes years of complex biological data in hours and finds patterns humans miss until it is too late

COVID-19 was declared a pandemic in March 2020. The first cases appeared in November 2019. Four months where the right AI running on the right data could have changed everything

GPT-Rosalind is now that AI. Running on that data. Inside the institutions that make those decisions

The same company building the most powerful AI in history just became part of the infrastructure layer of global pandemic response. They decide who gets access. They define what counts as a biological threat. They control what the AI sees and what it flags

That could be the best thing that ever happened to humanity. An AI that detects the next COVID before it leaves the first city. That develops a vaccine in days instead of years. That saves millions of lives before most people know a threat exists

Or it raises a question nobody wants to ask

When the AI monitoring global biological threats is owned by a private company - who exactly controls that information. And when they choose to share it.
COVID changed everything. The next one will too.
The only difference is this time an AI will know first.
And it answers to OpenAI

[Chonsy @ChonsyAI]

@NxtofAI They will handle it much faster

[Chonsy @ChonsyAI]

NASA built an AI chip that allows spacecraft make decisions without humans. 500 times faster than anything in space today

A Mars spacecraft currently waits 45 minutes for a human decision. This AI chip makes waiting obsolete. Testing began in February 2026

Here's how it works:

Every spacecraft ever built has one fatal flaw. It cannot think for itself

A sensor fails, an unexpected crater appears during landing, a radiation spike hits mid-mission - the spacecraft cannot react. It sends a signal and waits 45 minutes for a respond one way. 90 minutes round trip

By the time the answer arrives the landing has already failed - the mission is already lost

PIC64-HPSC ends 60 years of that problem:

> It fits in the palm of your hand
> Radiation hardened
> Survives temperature extremes that would destroy any consumer chip
> Performs 500 times beyond anything currently flying in space

Not 2 or 10 times, but 500

With this chip a spacecraft runs AI directly onboard. Processes sensor data in real time. Makes navigation decisions without asking Earth. Responds to emergencies in milliseconds instead of minutes

The first test at JPL began February 2026. The engineers marked the moment with an email titled "Hello Universe" - a nod to Hello World, the first message in computing history. They knew exactly what they were starting

This is what changes with one chip

A Mars rover that decides for itself when to detour around a rock field. A deep space probe that adjusts its own trajectory without permission. A lunar habitat that handles emergencies before Earth even knows they happened

For 60 years humans made every decision for every spacecraft that left Earth. Every single one

That era just ended

And the chip that ended it fits in your hand

[Chonsy @ChonsyAI]

A humanoid robot just entered a war zone for the first time in history. It is already on the frontline in Ukraine

The company’s co-founder said it plainly. „We think there is a moral imperative to put robots into war instead of soldiers.“ Two units are already there. 50,000 planned by 2027

Humanoid robots have been folding laundry, sorting packages, building cars

Nobody talked about this:

In February 2026 a US startup called Foundation sent two Phantom MK-1 robots to Ukraine. Built specifically for combat. The first humanoid robots ever deployed to an active war zone

They went straight to the frontline for reconnaissance

The Phantom MK-1 wasn’t designed for a warehouse. It was designed for war. Built to operate in environments too dangerous for humans, carry weapons, navigate battlefield terrain, perform reconnaissance under fire

The long term goal is that a robot can use any weapon a human soldier can

The US Army, Navy and Air Force already agreed. $24 million invested before a single unit shipped

And Ukraine was the obvious choice for the debut

Robotic ground systems were used in more than 7,000 operations in January 2026 alone in Ukraine. Drones. Unmanned vehicles. AI-guided systems. Ukraine has become the world’s largest real time testing ground for military technology. Every defense startup wants data from a real war

Co-founder Mike LeBlanc - a former US Marine Corps pilot described what he saw in Ukraine as „truly shocking.“ „This is a full-scale war of robots where the robot is the primary fighter and humans only provide support.“

Foundation is already building the Phantom MK-2. Waterproof. Better battery. 80kg payload. Production target is a 50,000 units by 2027

And one more detail nobody is reportin.

Eric Trump - son of the US President, recently joined Foundation as chief strategy adviser.  He was an investor before becoming an adviser. The most powerful political family in America is now directly involved in the company building humanoid combat robots

Now the machine is the soldier

And it is already there

[Chonsy @ChonsyAI]

They built a humanoid robot in 6 months. Renting it out for $10 an hour. Google DeepMind already have interest in it

Just a working humanoid robot built in 6 months that costs less per hour than the average warehouse worker

Here is what nobody is talking about

@thefinalcompany
November 2025. Eden Robotics is founded. Four people. No funding announcement. No press release. Just a mission - build the OS for the physical intelligence age

January 2026. Google DeepMind requests access to their data for a potential collaboration. The company is 40 days old

May 2026. EDEN 1 is a fully operational humanoid robot which is completely designed, manufactured and operational. Six months from founding to working hardware

Most robotics companies take years to get here

Eden Robotics did it in six months with four people

Here is what EDEN 1 actually is:

5 foot 6. 30 pound payload. 20 hours of battery. Autonomous charging. Runs on Theta OS - a general purpose operating system that works on any robot hardware from warehouse floors to the lunar surface. The same technology originally designed for satellite and lunar robots now powers a humanoid sorting packages in a factory

And here is the part that changes everything

You do not buy it. You rent it. $10 an hour. $800 a month includes 80 hours. $0 Setup fee. 30 day free trial with full refund if it does not meet expectations

The math is very simple:

One EDEN 1 costs $10 an hour. One warehouse worker costs $25 to $35 an hour fully loaded. One robot works 20 hours straight without a break. One human works 8

Eden Robotics was founded six months ago

Google DeepMind noticed in month two. The rest of the industry has not noticed yet

That window is closing fast

[Chonsy @ChonsyAI]

Before firing 8,000 people Meta tracked their every keystroke to train the AI that would replace them

The tool recorded mouse movements and keystrokes on every work computer. The data trained AI agents capable of doing those same jobs and then 8,000 people got their layoff emails

Zuckerberg had a very busy May. Nobody was supposed to notice

In 2022 he fired 21,000 people and apologized. "I got this wrong. I take responsibility for that." The stock was in free fall. He was contrite and called it a mistake

This time there was no apology

Just 8,000 jobs cut and 6,000 open roles cancelled. "All part of our continued effort to run the company more efficiently.» Without explanation and accountability

Because this time it is not a mistake

While the layoffs were being processed Meta quietly acquired Assured Robot Intelligence - a startup building AI brains for humanoid robots. The deal closed May 1. The layoffs started May 20. Three weeks apart

And then came the detail that nobody reported

Before firing 8,000 people Meta installed a tracking tool on every employee's work computer. Mouse movements. Keystrokes. Every action recorded. The data goes straight into AI training models that will be used to build digital agents capable of doing those same jobs

Meta was training its replacement workers on its existing ones

Employees called it dystopian - created a petition - called it nonconsensual data extraction - leadership kept the tool running

The CFO said during earnings that Meta doesn’t "really know what the optimal size of the company will be in the future." Translation - more cuts are coming. August, then later in the year. The number has no floor

110,000 tech workers lost their jobs in 2026 so far. At this pace it approaches the 2023 record of 260,000

One executive search strategist put it plainly. "Now the world understands that jobs are being replaced by machines. And if you're not doing that shareholders are getting upset."

The stock jumped on the layoff news

In 2022 Zuckerberg apologized for firing people. In 2026 he tracks their keystrokes to train the AI that replaces them. Then fires them anyway

No apology this time. Shareholders got what they wanted

[Chonsy @ChonsyAI]

The biggest day in AI history just happened. Most people missed two of the three stories

Three things happened on May 22 that had never happened before. All at once. If you only read about the OpenAI IPO you missed the two stories that actually matter more

Here is what they missed:

The same morning Anthropic posted its first profit in company history. $10.9 billion in revenue. Five years of losses. Gone. The AI that developers chose over every competitor just became a real business

And then the SpaceX S-1 dropped

Inside a 400 page filing - one line. Anthropic pays SpaceX $1.25 billion every single month for compute. $15 billion a year. Through 2029. The rocket company is now the backbone of the AI that just turned profitable

The math nobody ran:

OpenAI - $1 trillion valuation. Zero profit. Filing for cash it desperately needs. Anthropic - $900 billion valuation. First profit ever. Paying $15 billion a year to train the model that is beating OpenAI. SpaceX - $1.75 trillion. Making $15 billion a year from the company everyone counted out

Three companies in one morning
The AI race just reshuffled in public

The company that was supposed to lose - became profitable. The company that was supposed to win - filed for an IPO because it is running out of money. The rocket company is getting rich from both of them

Nobody wrote that story. Until now

[Chonsy @ChonsyAI]

@anis_neyo using powerlines as a charging network is genius

[Chonsy @ChonsyAI]

$1,000,000,000,000 valuation and $0 profit. The biggest IPO in history just dropped and Wall Street is lining up

Loses $1.22 on every dollar earned and burns cash faster than any company alive
Goldman Sachs said yes, Morgan Stanley said yes
Everyone wants in and nobody asked why

Yesterday OpenAI filed a confidential S-1 with the SEC. Targeting a public listing in September 2026. Valuation between $852 billion and $1 trillion

OpenAI has never made a profit, not one quarter

In Q1 2026 it lost $1.22 for every $1 of revenue. Full year 2025 revenue was approximately $13 billion. Losses were larger. The company raised $180 billion from investors and burned through most of it building the infrastructure to run the product 500 million people open every week

The product that autocompletes your emails. Writes your code. Answers your questions at 2am

And here is the part that makes this truly unprecedented:

Faster than any company in history
This is not a startup gambling on a future that might arrive. The future already arrived. ChatGPT already works. The users are already there. Revenue grew from $1 billion annually in 2023 to $24 billion annualized today

The problem is the costs grow just as fast. Every answer costs compute, every user costs infrastructure, every new model costs billions to train. The more people use it the more it costs to run

But here is the bet Wall Street is making:

AI becomes the operating system of the global economy. The company that built the product everyone already uses wins everything
Profit follows dominance - Dominance is already here

Three AI companies. Three trillion dollar valuations in six months:
OpenAI isn't alone. Anthropic is targeting a $900 billion IPO in October. SpaceX already filed with xAI inside.

Google went public in 2004 losing money. Amazon lost money for a decade. Both became the most valuable companies on earth

September 2026. The biggest IPO in history

[Chonsy @ChonsyAI]

Genesis AI just solved the hardest problem in robotics. A robot that manipulates the world like a human

Backed by $105M and the former CEO of Google

It cooks a 20 step meal, solves a Rubik's Cube in mid-air, plays piano, conducts lab experiments. With one hand. This changes everything

Here is what nobody is talking about:

For 50 years robots could see, they could think, they could move from point A to point B
The one thing they could never do was touch the world the way humans do

Pick up an egg without breaking it, thread a wire through a tight harness. Transfer liquid between containers without spilling. Play a chord on a piano. These tasks require something robots have never had.

Human-level dexterity

Genesis AI just built it:
GENE-26.5 is the first AI brain that gives robots human-level physical manipulation. Not approximately human. Not close enough. Human level. The hand mirrors human anatomy almost bone for bone. Every joint. Every degree of freedom. Every tactile sensor mapped 1 to 1 between the human hand and the robot hand

The demonstrations aren't marketing

A robot cooking a 20 step meal from scratch. Solving a Rubik's Cube continuously in mid-air requiring coordinated high speed reasoning. Playing piano at human level performing an ultra fast complex composition. Pipetting liquid in a lab with delicate instrumentation. Bundling wire harnesses — one of the hardest tasks in electronics manufacturing. Grabbing four objects of different sizes simultaneously with one hand

Every single task done by the same brain. The same hand

And here is the part that makes this commercially terrifying

The data glove that trains the system costs 100 times less than typical alternatives. Collects five times more usable training data. Which means Genesis AI can train faster, cheaper and at a scale nobody else can match

Eric Schmidt, former CEO of Google, called it a paradigm shift. Vinod Khosla said it is changing the trajectory of robotics. $105 million seed round. The largest in French robotics history

The general purpose robot is coming next

And when it arrives it will have the most capable hands ever built

[Chonsy @ChonsyAI]

@L1vsun The course always ends with a recommendation for the next one. By design:)

[Chonsy @ChonsyAI]

@ty_westenbroek The visualization alone is worth clicking!

[Chonsy @ChonsyAI]

Google just killed every video editor with one model. It's called Gemini Omni and it launched today

It understands gravity, light, motion, physics. Give it anything and it generates video that looks like a production team made it. One prompt. One model. Launched today. GAME OVER

Here is what nobody is saying out loud:

3,000,000 people worldwide get paid to edit video. In studios, as freelancers, on Fiverr, on Upwork, in agencies. They all went to work today not knowing this was coming

Google didn't send a warning, didn't announce a transition period, didn't offer retraining. They just opened Google I/O, put Demis Hassabis on stage, and showed a photo turning into a finished cinematic video in real time

The audience went silent.

Here is what Gemini Omni actually does:
>You give it a photo
>A voice note
>A sentence.
It gives you back a finished video with correct lighting, correct physics, correct motion. It understands how rain hits pavement, how a human being moves through a room.
You get finished product that looks like a senior editor spent a day on it

Adobe spent 40 years building tools for this workflow and Google collapsed it into a single model today

One Gemini subscription $20 a month. Gemini Omni render - seconds. Human edit - a full day
Available right now to 900 million users worldwide

The moment that math lands with enough clients the industry doesn't slow down. It stops

And the only people left standing are the ones who knows how to direct these models before everyone else does

Everyone else found out this morning. Most of them just don't know it yet

[Chonsy @ChonsyAI]

@0xChaseTM Really useful information, appreciate it! Didn't even know about "A loss hurts your brain twice as hard as the same win feels good"

[Chonsy @ChonsyAI]

@matrixrobots Looks funny

[Chonsy @ChonsyAI]

@graftoverflow leopard print looks funny I would take him haha

[Chonsy @ChonsyAI]

@jaimandal_ Congrats!!

[Chonsy @ChonsyAI]

A Chinese robotics company just shipped more humanoid robots than Tesla, Figure AI and Agility combined

The robotics race has a new leader and it is not American
Its name is @UnitreeRobotics
Revenue up 335% in 2025. First profitable year. 5,500 robots shipped and filing for a $610M IPO

The price of their humanoid fell from $85,000 in 2023 to $25,000 in 2025. Target is 75,000 units per year within five years. 20,000 this year alone

Wang Xingxing left DJI in 2016 to build robots nobody thought would sell. His first product was a quadruped he designed for his master's thesis. It went viral and Investors showed up

The goal is simple:
Robots that genuinely work. Not perform. Not dance at a gala. Work and liberate productivity. Handle the tasks humans don't want to do

Ten years later Unitree is the largest humanoid robot manufacturer on earth

Unitree's three biggest American competitors are Tesla, Figure AI and Agility. Each sold around 150 units. Unitree sold more than all of them combined. 32% of the entire global humanoid market shipped in 2025

And it did it profitably. $250M revenue. $90M net profit. 60% gross margins. The first profitable year in humanoid robotics history

Unitree isn't just a hardware company, the AI story is what most people miss, because their own world model UnifoLM-WMA-0 is already open sourced

Nearly half of the $610M IPO proceeds go directly into training AI models. It lets robots predict and adapt to physical environments, spot a pothole while walking and adjust before you fall

Wang's own words on where embodied AI stands right now: "Current robot AI resembles the stage before ChatGPT emerged. A critical breakthrough is coming in 2-5 years."

Elon Musk saw Unitree's latest mecha robot and commented one word "Cool."

Morgan Stanley said China's early lead in humanoid robots could drive the next phase of global manufacturing dominance