I Ask What They Saw Through the Telescope

Brad Lightcap has seen things

5:29

Brad Lightcap has seen things...Not in the Blade Runner sense. Probably.

But for eight years, he has occupied one of the strangest seats in modern technology: close enough to OpenAI's research frontier to understand what was coming, while simultaneously responsible for turning that capability into an actual company.

He saw the boardroom war. He saw Microsoft become both rocket fuel and constraint. He saw the arguments over what AGI even means move from philosophy into contract language. He saw enterprise demand arrive faster than the organization could comfortably absorb it. And he has seen things that we haven't seen yet.

Brad Lightcap joined OpenAI in 2018. He became CFO. Then COO.

He ran operations. He ran the Startup Fund. He was the guy who had to make the business actually work while Sam Altman was out front painting the vision.

When Altman got fired by his own board in November 2023 and then un-fired four days later, Lightcap was inside the building for all of it. When Microsoft went from investor to life support to potential cage, he was in those rooms. When OpenAI's enterprise business started exploding and the org couldn't figure out whether it was a research lab or a trillion-dollar company, he was the one getting the operational bill.

He was, arguably, the person with the widest operational field of view inside the company. And now he's leaving.

Not retiring. Not joining another behemoth. He said he's going to "start something new."

So when somebody with that field of view walks away from what may become one of the most valuable companies ever created and says he's going to start something new, my ears go up.

Not because OpenAI is collapsing.

Because Brad Lightcap apparently believes there is something outside OpenAI worth leaving OpenAI to build.

That's the part I can't stop thinking about.

I've been writing about the fault lines inside OpenAI for years.

In 2023, when the Altman board explosion happened, I was looking at it as an ideological struggle. Not a personality conflict. A genuine fight over speed, control, and who gets to decide what responsible scaling actually means. The surface story was drama. The underneath story was about the gap between people who wanted to move faster and people who wanted to build more guardrails, and the fact that nobody could agree on which side was the idealistic one.

In 2024, I was asking whether people inside OpenAI believed they had effectively crossed the AGI threshold. Not as a thought experiment. As a practical question. Because the Microsoft agreements actually hinged on the definition. If OpenAI declared AGI, certain contractual obligations to Microsoft were supposed to change. Which meant the philosophical question about what counts as general intelligence had become a financial question with billions of dollars attached to the answer. The people inside that building had to think about that differently than the rest of us speculating from the outside.

And in 2025, I was looking at OpenAI trying to escape the structural limitations of its Microsoft relationship because enterprise demand was pulling it into a much larger multicloud world. The gilded cage. Microsoft was the thing that made OpenAI's scale possible, and also the thing that made certain kinds of growth structurally very difficult. The enterprise customers didn't want to be locked into one cloud. OpenAI needed to be free to meet them where they were. But the deal said what the deal said.

Three chapters...The board war...The AGI contract problem...The Microsoft cage.

Brad Lightcap lived through all three. Not as an observer. As the guy running operations while those fights were happening. And now he's walking out the door.

This is the point where I should probably speculate about what he's building. I'm not going to.

Not because I don't have theories. I do. You can construct a plausible argument that somebody who spent eight years watching the gap between what frontier models can do and what enterprise customers actually need is going to build something in that gap. You could argue that somebody who watched the Microsoft dependency problem from the inside might build something designed to avoid that trap. You could argue that somebody who managed the OpenAI Startup Fund and saw hundreds of portfolio companies might have developed a very specific thesis about where the value migrates next.

Those are all reasonable. They're also all guesses.

What I know is this.

Brad Lightcap spent eight years standing closer to the event horizon than almost anyone on Earth. He was there before ChatGPT. He was there during ChatGPT. He was there for every iteration of the definition of what they're building and why it matters and who controls it.

He watched the organization transform from a weird research nonprofit into something approaching an industrial machine. And then he decided the thing worth building next is somewhere else.

I don't know what Brad Lightcap is building.

That's precisely why I'm paying attention.

When people like that leave the observatory and start building something of their own, I don't immediately ask whether the observatory is in trouble. I ask what they saw through the telescope.

---

Rich Washburn is a technologist and strategist working at the intersection of AI, infrastructure, and capital. He is Managing Partner and Chief AI Officer at Eliakim Capital.