The short version
I founded 8GI as a nonprofit foundation because I believed a few specific things about where the AI economy was heading. That compute would get rationed. That inequality would compound. That hardware dependency is a trap. That the safety narrative gets weaponised into permission for incumbency.
For two years I wrote the argument. In AI Bill submissions in Ireland and South Africa. In policy briefs. In open letters. In this feed.
Then Sam Altman and Greg Brockman sat down on the Core Memory podcast with Ashley Vance and Kylie Robinson, and between them, in one episode, said the quiet part out loud on every single one of those things.
This post is 8GI's case for its own existence, told in their voices, not mine. Five pillars. Five quotes. All verbatim from the tape.
Pillar 1. Compute is the new property qualification.
Greg Brockman, President of OpenAI, on Core Memory, April 2026:
"AI is really opportunity. AI is opportunity for everyone if you have access. If you have compute. If you don't have compute, you can't."
Read the hinges. If you have access. If you have compute. If you don't have compute, you can't.
Opportunity in the AI economy, in the words of the President of OpenAI, is conditional on compute. Not on effort. Not on education. Not on talent. Compute.
What 8GI does about it. Our open-source agent kernel runs local by default on the machine you already own. No API key to start. No subscription to clear the door. Free is not the introductory tier. Free is the architecture.
Pillar 2. The business model is rent.
Greg, thirty minutes later in the same episode:
"Compute for us is not a cost center, it's a profit center. We rent or buy compute and then we resell it at a margin. And as long as we have some positive margin on it, then it's scalable. Because the demand is just unlimited."
Compute is a profit centre. The business is buying compute and reselling it at a margin. Greg is not ashamed of saying this. He is describing the economics correctly. That is his job.
My job is to notice that a business whose scalability depends on positive margin on compute cannot, by construction, ever deliver compute at cost. Which makes Sam Altman's line from three minutes earlier in the same podcast structurally impossible:
"Kind of no matter what, I think everyone should want much more compute, much more infrastructure and the cheapest possible access to AI because otherwise I think you really exacerbate inequality if there's a limited amount of this and the price goes up because of supply and demand and only the rich people have it."
Cheapest possible access is a cost structure. A margin-priced resale is a cost-plus structure. They are not the same thing.
What 8GI does about it. 8GI is a nonprofit foundation. Our equity is locked inside the foundation itself, not in a founder or an investor. Our code ships Apache 2.0. When we spend on compute, it is a cost, not a profit centre. No margin to extract, because there is nobody to extract it for.
Pillar 3. The preferred outcome is ten trillionaires.
Sam Altman, same podcast, describing what he calls the first of three futures:
"The floor comes way up. Everybody gets subjectively like 10 times richer. But in that world these people who really learn how to use agents and get a lot of compute together, maybe we have 10 trillionaires. The floor comes way up, but because this is a lever that people can really use, the people who already started rich and have access to a lot of compute, inequality gets worse."
A minute later:
"We think it's obvious, Greg, and I think it's obvious that people should prefer the first."
The first. The ten-trillionaire world. With worsening inequality. That is the preferred world, in the mouth of the CEO of OpenAI, on a major podcast, in April 2026.
What 8GI aims to do about it. We envision Universal Basic AI. A citizen entitlement, denominated in inference, attached to the same constitutional logic as education or healthcare. Not charity. A floor. Markets do not build floors. Foundations do. We have UBAI in our Irish AI Bill submission, and I am submitting to some more AI policy makers globally in the coming weeks.
Pillar 4. The hardware is somebody else's.
Still the same podcast, Sam on the US industrial base:
"The US is extremely behind here."
"We can't even make an actuator."
"The US has no credible plan other than the plan we had of AI plus robotics."
The entire United States strategy for staying competitive in physical production is one lab's bet on a stack it has not shipped yet. Any country that picks "American frontier AI" as its infrastructure play is picking a stack whose hardware layer is, by the admission of the CEO of the leading lab, not ready, not buildable at home, and dependent on supply chains the US does not control.
What 8GI does about it. Sovereign and local-first by architecture. The 8gent kernel runs on the machine in front of you, against models you can download, with no API call required to function. If the undersea cable is cut, if the data centre is offline, if the API tier is repriced, the agent still works. This is not a political position. It is an engineering one.
Pillar 5. The safety story sells bomb shelters.
Near the end of the same episode, Sam reflects on how the industry has marketed itself:
"I think we built a bomb shelter for $100 million."
He is talking, self-deprecatingly, about the way the AI industry, OpenAI included, has marketed through existential fear. "We have to be the ones who build it, because otherwise someone worse will." This is the oldest pitch on earth, and it is how the frontier labs have written policy, acquired permits, and absorbed advisory seats in governments across the Western world.
I am not saying AI safety does not matter. It matters enormously. I am saying that "we built the bomb shelter" is not safety. It is a sales pitch with a cross on the door.
What 8GI does about it. We ship an agent kernel you can run on your own laptop today, and if we are wrong about the future, the worst thing we have done is give away some open-source code. The work speaks for itself.
What that adds up to.
Five pillars, each named by the co-founders of OpenAI in a single episode of a single podcast, in a single week in April 2026.
- Opportunity is conditional on compute. (Greg)
- The business is reselling compute at a margin. (Greg)
- The preferred world is ten trillionaires and worsening inequality. (Sam)
- The US hardware plan is the AI plan, and the AI plan is not ready. (Sam)
- The industry's marketing is a bomb shelter for $100 million. (Sam, on himself)
These are not my claims. They are the claims of the people with more commercial interest in the current trajectory than anyone else on earth.
8GI Foundation is the answer, shaped as follows.
- Against pillar one. Free and local by default.
- Against pillar two. Nonprofit. No equity extraction.
- Against pillar three. Universal Basic AI, in the Irish AI Bill submission.
- Against pillar four. Sovereign and local-first.
- Against pillar five. No fear marketing.
The close.
I did not start 8GI because I think Sam and Greg are villains. I don't. I think they are sincere in every line I quoted. That is what makes the problem structural, and not personal. A frontier lab operating as a commercial entity has to sell compute at a margin or it cannot fund the next training run. Everything downstream of that follows from the structure.
8GI exists because someone has to build the counter-structure.
The K is a shape. The 8 is a choice. Sam and Greg, on Core Memory, named the K. 8GI is how we build the 8.
Sources: Core Memory podcast, "Sam Altman and Greg Brockman on 10 years of OpenAI," released April 2026. Timestamps: 39:05 to 39:42 (Sam, ten trillionaires), 42:10 to 42:26 (Sam, cheapest possible access), 42:50 to 43:10 (Greg, if you don't have compute), 46:00 and 46:20 and 47:06 and 47:46 (Sam, US hardware), 1:00:23 (Greg, compute is a profit centre), 1:04:50 to 1:05:24 (Sam, bomb shelter). Excerpts verbatim. Companion piece: "If you don't have compute, you can't." at 8gi.org/blog/compute-you-cant.
By James Spalding · Dublin · 8GI