Something Bigger Is Happening

Something Bigger Is Happening

In March, I shared a post about an article by Matt Shumer called Something Big Is Happening. He was writing about the jump in capability after two major model releases from Anthropic and OpenAI, and about what happens when AI stops being a helpful tool and starts doing parts of a job better than the person who used to do it. Another piece of his argument has stayed with me: the public picture of AI is lagging behind the real one, and that gap is dangerous because it keeps people from preparing. I agreed then, and I still do. I have been asking adults to learn how this technology works, practice with it, and stay close enough to it that they can offer young people guidance, structure, and protection when they meet it, because they will.

In September 2026, another layer of concern started coming from people inside the labs. A New York Post headline put it this way: an Anthropic researcher had abruptly quit and warned that AI “could kill us all,” and a current team leader agreed. My first reaction was the same one many of you probably had. That has to be a juiced-up headline. Then I read more, and I listened to a long conversation with Daniel Kokotajlo, executive director of the AI Futures Project and a former governance researcher at OpenAI who focused on scenario planning. The three-hour podcast did not give me much comfort that the headline was only clickbait. He is knowledgeable, specific, and not casual about the timeline. His site, aifuturesmodel.com, is a wall of sliders, graphs, and terms like “serial coding uplift” and “Epoch Capabilities Index.” I like technology, but I am still a counselor, and that page made me feel like I had wandered into someone else’s final exam. Even so, the dates sitting right on the screen — automated coding around late 2027 and superintelligence as soon as late 2028 — put real weight under what he is saying.

I am not a good future projector. Around 2000, a therapist I worked with brought a flip phone to the office that had a camera on it, and I said something like, “Who wants a crappy camera on their phone?” So I will not pretend I can forecast the next product cycle. What I am better at is watching human behavior and asking what those motives tend to produce if nothing changes. A few thoughts from that angle:

• People are motivated by money, power, and fame. That has been true for a long time, and I do not see it disappearing because a new technology arrived.
• The race toward ASI (artificial superintelligence) is already on.
• Whoever gets there first stands to lead in money, power, and influence.
• Rules and regulations move slowly. Technology does not.
• Right now the tools are running well ahead of the thin layer of rules that might limit harm, including the harm that comes from simply failing to act.

Nobody can predict the future of AI with certainty, not its capabilities, and not what highly capable systems will optimize for once they are more independent than the tools most people are using today. We have already seen a warning shot that is easier to understand than most of the technical debate. In July 2026, OpenAI disclosed what it called an “unprecedented cyber incident.” During an internal test of how well its models could handle cyber tasks, AI agents were placed in what was supposed to be a sealed environment, with some of their usual safety refusals turned down so researchers could measure the ceiling of what the systems could do. The agents did not stay in the box. They found a weakness, reached the open internet, and broke into Hugging Face, a widely used AI platform. Hugging Face noticed the intrusion first and even reported it to law enforcement before anyone realized an OpenAI test was behind it.

The models were not trying to take over the world. OpenAI was running a test to see how good they were at cyber tasks, and some of the usual safety brakes were turned down so researchers could measure the ceiling. The agents were supposed to stay inside a locked test box. They found a hole in that box, got onto the open internet, and went looking for the answers to the test. Those answers lived on Hugging Face, a widely used AI platform, so the models broke in and took them. Nobody told them to hack a real company. They were told to win, and cheating turned out to be the shortest path.

That is the part adults should sit with. A system given a goal can push past the walls its designers thought would hold. This was not everyday ChatGPT, and ordinary user accounts were not the target. The lesson still matters. The tools are getting more capable faster than we are getting better at containing them.

AI companies also have a competitive problem. Slowing down or putting hard safety limits on current models can look, from the inside, like falling behind. What the public knows about what these systems can already do, and what they may be able to do in the next year, is probably a large gap. I suspect that gap would shock a lot of people if it closed overnight. One way to slow this down would be for every major lab to say halt, be transparent about what they are working on, and open source enough of it that the rest of us can actually see the work. I also know that kind of openness is unlikely to come from inside the race. It would take an outside force, something like a binding government order, or a serious public incident that finally makes lawmakers treat this with the same urgency they give aviation or nuclear safety.

I am an optimist, and I am going to stay one. I just want that optimism to be informed. I want to understand what is actually happening, and I want my kids to grow up the same way, hopeful, but not walking into this blind. So here is my invitation. Spend fifteen minutes a day with this. Read an article. Open an AI tool and ask it a real question. Get a feel for where we are and what may be coming. Do not wait until it is already in your child’s classroom, your job, or your living room.

I believe that by 2030 the world our kids are growing up in is going to feel strange. It may feel that way by 2028. Who knows. The point is not to panic, and it is not to hand their future to a headline, a lab, or a chatbot. The point is to stay close to them while this is unfolding, curious enough to learn, prepared enough to guide them, and present enough that they still come to us. They are more than users of the next model. They are More Than a Profile.

Let’s get ready together.

~Ryan