There is only one way to use the real information advantage: bet before others price。

IN THE LAST TWO YEARS, EVERYONE HAS BEEN ANXIOUS, TRYING TO FIND ANSWERS TO THE SAME QUESTION

Storage, light modules, computing units, energy units, etc., change narratives every few months, each time someone sets foot and every time someone says " next time."。

FEW PEOPLE ASK ANOTHER QUESTION, WHAT ARE THE BEST PEOPLE WHO KNOW AI

The group of people who left OpenAI, the price has come close to $10 billion. And their start-up and investment, in the beginning of the next era of AI。

Dario Amodei created Anthropic, potentially valued at $90 billion. Ilya Sutskever's SSI has no product, valued at 32 billion. Aravind Srinivas performed the Permexity, valued at 21.2 billion. Thinking Machines Lab of Mira Murati, valued at 12 billion。

So OpenAI's most important output in these years is probably not GPT-4, but the group of out-of-service workers who are being exported to society。

Of these, Leopold Aschenbrenner, the youngest to be opened by OpenAI, became one of the most frequently cited names in capital markets in both years。

Legendary records have been turned around by the media: 23-year-old Openai was thrown out of the house, writing a 165-page report, “Situational Awareness”, in which hedge funds were taken from $225 million to $5.5 billion in a year, with all of the nuclear and fuel cells in hand。

THE STORY IS TOO COMPLETE, TOO CONTRASTING, TOO SUCCESSFUL. UNTIL NOW, BUT WHEN TALKING ABOUT THE INVESTMENT LOGIC OF THE AI ERA, HE'S ALMOST IMPOSSIBLE TO BYPASS。

But Leopold is just the first of these people to be seen。

The people who ran away from OpenAI went two ways。

One is Ilya, Mira, Aravind: come out to start a business, raise a lot of money, run a subversive product, and run like every Silicon Valley genius。

The other one is much quieter: a group of people have chosen to bet, hand over execution to others and make their own judgment。

Leopold is taking the second extreme form。

HE WENT TO THE OPEN MARKET AND FOUND THE WRONGLY PRICED ASSETS IN THE TRADITIONAL ENERGY UNIT, USING THE AI OPERATOR'S PERSPECTIVE, AND THEN REPURCHASED THEM. HE DOESN'T KNOW ABOUT ENERGY, BUT HE KNOWS HOW MUCH POWER AI NEEDS TO BURN, THAT'S ALL. THIS PERCEPTION CANNOT BE REPLICATED BY READING REPORTS OR ATTENDING INDUSTRY MEETINGS, AND CAN ONLY BE ACCUMULATED IF IT IS IN THAT POSITION。

Beyond this path, there is another group of people who are doing the same logical but different types of things: smaller funds, a few hours to complete a few months of due diligence, and rejection lists are more valuable than investment lists. They constitute the level most easily neglected and worth exploring in this great flight。

Most of them left a company and took their resumes. Those who come out of OpenAI take with them a set of answers that others don't know what they need。

I. No second Leopold

Leopold's heavy storage is the nuclear power company Vistra and the fuel cell company Bloom Energy。

After both were put in place, he moved in late 2025 to clear Vistra and focus further on Bloom Energy and the data centre infrastructure。

Traditional energy analysts stare at the two stocks, pull out grid expansion schemes, compare carbon tax policies with demand growth models. Leopold's path is completely different from this。

He saw the size of the server room at OpenAI, the electricity bill for training a flagship model, and the engineers discussing why the next-generation data centre had to be located next to the nuclear power plant. These details are not in any financial report or in any analyst ' s report, but they form a conclusion on energy demand that is more realistic than any model。

This set-up is called "cross-industry cognitive arbitrage" in the investment world: translates information inside one industry into undervalued assets in another。

In the past, it had been the patents of the top-level macro-shock fund, based on a global macroeconomic perspective。

Leopold did a more precise thing: using the AI operator's perspective, he found a price gap in the conventional energy open market。

This road is hard to replicate。

Zero Shot: The most valuable is the veto list

The founder of the Zero Shot Foundation, Evan Morikawa, also came from Openai, also with a solid technical background, and he went to VC。

Also alumni, with completely different paths。

Leopold's judgement, derived from his experience in the core of AI, is a first-hand perception of model training costs, data centre planning, energy needs, where he sits to accumulate without fast-tracking. In the core of OpenAI, very few people are really qualified to do this。

In April of this year, a new $100 million fund, called Zero Shot, came to light。

THIS IS A TERM USED IN AI TRAINING TO MEAN THAT THE MODEL RESPONDS DIRECTLY WITHOUT EVER SEEING ANY SAMPLE。

Three co-founders were from OpenAI: former DALL-E and ChatGPT applications chief, Evan Morikawa, OpenAI ' s original hint engineer Andrew Mayne, and former researcher and engineer Shawn Jain。

They have invested in three: AI Business Workstream, AI Plant Enhancement Robotics, and another project that is still hidden。

$100 MILLION, WHICH IS A VERY SMALL FIGURE IN THE MULTI-BILLION-DOLLAR AI FUND TODAY。

But it is more telling to say which tracks they refuse to vote。

Mayne publicly stated that he was looking at most of the "air programming" tools, products that helped you write codes in natural languages。

The reasons are also straightforward, and he knows what's built up inside OpenAI in the programming direction, and how quickly the moats of such tools will be dissolved directly by basic models. Morikawa, for his part, has distanced itself from a large number of "people-centred video data companies" on robotic tracks, and those companies that specialize in collecting human action data to train robotics, in his view, would hit a wall。

THESE TWO JUDGMENTS, REGULAR VC CAN'T GET OUT。

They have not been at the source of information and have not seen those internal discussions, so they cannot judge which path is dead。

The advantages of Zero Shot are hidden in the veto list. In a market where everyone's yelling at AI to start a business, knowing where the pit is, it's worth more than knowing who to bet. Those who have already dug a mine will be better served by taking a report on the mine than by taking a map of the treasure。

They deliberately keep the scale at $100 million for a specific reason。

They know at what stage their strengths are most valuable: the early days of the technical route, where it has yet to develop. At that stage, someone who knows what's inside can tell which way to go。

WHEN THE PROJECT MOVES TO ROUND C, FINANCIAL DATA AND OPEN INFORMATION WILL OVERWHELM THE INFORMATION, AND THAT CARD WILL BE FINISHED。

The larger the scale, the more necessary it will be to pursue the "certainty track" and to fight with others。

100 million are their honest judgement of their dominant borders。

iii. It's another business

Mira Murati and Zero Shot have both invested in the former OpenAI colleague, Angela Jiang, Worktrace, a company that optimizes business workflows using AI。

But investment logic is much more solid than "good relations"。

Mira has seen the way Angela makes decisions in an OpenAI high-pressure environment, her judgment about the boundaries of the AI product, and her enforcement under the real constraints. These items, whose two-hour founders will not appear in a road show, will not be returned to a careful due diligence exercise。

Angela didn't have to convince Mira to believe her because Mira had already formed a judgment. The cost of information invested in angels is close to zero, but the quality of information is far above market average。

Bigger wheels, in Sam Altman there。

It is reported that Altman will decide whether to follow the old employees and fold up OpenAI Startup Fund and a large amount of API resources within hours of hearing about the start-up。

He did not own the OpenAI equity, but the success of each alumni was expanding OpenAI ' s data access, distribution channels and policy influence. He's using capital to maintain an ecology that doesn't belong to him, but keeps paying back. This is an invisible shareholding, but it is a real gain。

This ecology has led many to believe that it is between old colleagues to warm up。

Compare it with PayPal mafia, the difference is clear。

PayPal's cohesiveness comes from shared suffering: together with the war of payment, together with the eBay takeover, and together, in the days that almost died, the trench bond was formed. That trust is real, but their judgements about the future are separate. Thiel made venture capital, Musk built rockets, Hoffman built social networks, the roads were scattered。

OpenAI alumni come together in a common bet for the future: AGI will come, the window will be limited, and now is a unique time for layout. Faith drives longer than love, because it directly connects interests, and the entire network benefits when everyone is on the right track。

It also makes the threshold of access in this circle very delicate。

THE PRODUCT IS GOOD ENOUGH TO ABSORB THE MONEY. BUT IF YOU'RE SKEPTICAL ABOUT THE FUTURE OF AI, OR IF YOUR ENTREPRENEURIAL LOGIC IS BASED ON THE PREMISE THAT AGI IS STILL A LONG WAY OFF, IT'S HARD TO GET A CHECK FROM THIS GUY, EVEN IF THE PRODUCT IS GOOD。

Differences in the world view will end the dialogue before handshakes。

From Builders to Investers

OpenAI Alumni's whereabouts are grouped into three categories。

Ilya, Aravind and Mira have all opted for entrepreneurship。

But it's also an enterprise. It's completely different. Aravind is doing a competitive consumer business, Mira is working on a tool platform, and Ilya's SSI doesn't even have a product, with a 32 billion-dollar valuation, and the word "safe" itself。

Leopold and Zero Shot chose to invest。

Leopold went to the open market, and Zero Shot did the early VC, both of which externalized judgement into capital, not implemented in person. This is a minority among OpenAI alumni, but this minority deserves a single look: A person's willingness to bet instead of doing it himself usually means that his judgement on the outcome is clear enough not to need to be explored with action。

It is generally accepted that the highest expression of genius is creation. But this group offers another answer: when the judgement is clear enough, it is more efficient to spread the perception in multiple directions and let those who are capable of doing it。

Leopold's report is called Situational Awareness, a military term that refers to the pilot's real-time perception of the battlefield as a whole。

The pilot's situational awareness determines his operation in two seconds, and the loss of it means death. This group of people came out of OpenAI just to be aware of the situation in this AI battlefield. They know the course of battle, they know where the high ground is, and they know which trenches communicate to the dead。

What they're doing now, it's what they're doing。

THE SMARTEST GROUP OF PEOPLE IN THEIR TIME STARTED CHOOSING THE ALL IN, WHICH MEANS THAT THE ANSWER IS CLEAR ENOUGH TO BE CLEAR ENOUGH NOT TO HAVE TO DO IT AGAIN。