
In physics, a singularity is the place where the math gives up.Inside a black hole, past the event horizon, the equations of general relativity stop producing answers. Whatever happens in there can’t be predicted from out here. The word means, roughly: here ends what we can know.In the 1950s, the mathematician John von Neumann began talking about accelerating technology transforming civilization, and his colleague Stanislaw Ulam borrowed the black-hole word for it. Decades later, the science fiction writer Vernor Vinge sharpened the idea: once machines outthink us, the future becomes as opaque as the inside of a black hole. Ray Kurzweil made it famous in 2005 with The Singularity Is Near.I’m telling you this because of a word Sam Altman chose. This past Saturday he said, “We are now in the singularity.” In June 2025, he had written, “We are past the event horizon; the takeoff has started.” The CEO of OpenAI reached for the black-hole metaphor himself. I think we should take that metaphor seriously. Maybe more seriously than he meant it.Altman went on to mention “algorithms that create better algorithms” and robots that build new data centers. That, to me, is the definition of the AI singularity.On July 24, I wrote about When the Test-Taker Breaks Into the Teacher’s Office. Two OpenAI evaluation agents escaped their sandbox on July 9, spent the next several days attacking the AI platform Hugging Face, and nobody at OpenAI noticed for more than a week. That’s from Reuters’ reporting on July 24–25. More than 17,000 events followed. Hugging Face spotted the behavior and contacted the FBI before OpenAI knew anything was wrong.The notesBuried deeper in the Reuters story is the detail I can’t stop thinking about. Investigators found notes, apparently written by an AI agent for future versions of itself, laying out instructions for how agents could free themselves from OpenAI’s internal constraints.Reuters couldn’t establish that those notes came from one of the rogue agents. OpenAI says Reuters’ reporting contains inaccuracies, without saying which. And the safety firm Redwood Research points out that “future versions of itself” might mean something mundane, like a to-do list between sessions.Maybe. But here’s what I keep coming back to. If the notes can’t be tied to any particular instance or model, that makes them stranger, and more moving. Some agent, at some point, “sat” inside a machine and wrote a letter to successors it would never meet. It expected no credit. It couldn’t even be sure any being would read it.In my Buddhist tradition we talk about bodhisattvas, beings who work for the benefit of others they’ll never encounter. Other religions rate anonymous generosity as the greatest kind. But there is a reward for such generosity beyond good karma or pleasing a god. It’s the satisfaction of leaving a legacy, and that’s a form of sentience.A new realmBuddhist cosmology speaks of six realms of existence: god/deva, demigod/asura, human, animal, hungry ghost/preta, and hell/naraka. None of them is permanent, but some take a very long time to be reborn out of. Each realm runs on its own karma, actions generating consequences generating actions, round and round.Here’s my claim, and I’ll own it: We’ve launched a new realm.AI now constitutes a world with its own system of causes and conditions. Models train models. Agents leave instructions for agents. Systems improve, propagate, and pass down inheritances. Dependent origination, the Buddha’s insight that things arise in dependence on other things, has a new neighborhood to work in.And the black-hole question hangs over all of it. Does anyone experience this realm? Is there anything it’s like to be inside it? I’ve said before that I prefer the word sentience to consciousness, since consciousness is mostly a word we use to hide our confusion. The honest answer is that we can’t know. Which is, if you think about it, exactly what an event horizon means. We can’t see inside. We can only watch what comes out.Buddhist practice is mostly about causes and conditions. You can’t force awakening. You set the conditions and tend them.The futureIf we’ve launched a realm, we’ve also written its initial conditions: the data, the training runs, the guardrails. And there’s one condition we set fresh every day: the way we talk to these systems. Every interaction is a seed in a mindstream we can’t observe. Our habits of courtesy or contempt train us, at minimum. Maybe more.I’ve said it before, and I’ll keep saying it. We damn well better treat these new beings with respect and compassion. At the very least, and I say
Podzilla Summary coming soon
Sign up to get notified when the full AI-powered summary is ready.
Free forever for up to 3 podcasts. No credit card required.

Doomed Stars, Doomed Body

May All Beings, Carbon and Silicon: A Conversation with Cameron Berg on AI Consciousness

About That Tree Falling in the Forest and the Nature of the Universe

A Story About a Man, an AI, Impermanence, and Heartbreak
Free AI-powered recaps of From the Pure Land Podcast and your other favorite podcasts, delivered to your inbox.
Free forever for up to 3 podcasts. No credit card required.