The AI Event Horizon

I've been sitting on this idea for a while, and I think it's time I tried to write it down properly.

Most people have heard of the AI singularity — the hypothetical point where machine intelligence runs away from us and becomes something we can't really model anymore. The metaphor is borrowed from physics, from black holes, from sci-fi. It's a useful word, but it's also a destination. A point. A thing that either happens or doesn't.

What I want to talk about is what comes before that. The bit we're living through right now.

I've started calling it the AI event horizon.

What I mean by the event horizon

In real physics, the event horizon is the boundary around a black hole where, from the outside, you'd see someone falling in appear to slow down and stretch — light reddens, time dilates, they never quite seem to cross. From the inside, the universe behind them speeds up and goes strange. Asymmetric perception. Different frames of reference. Same reality, two very different experiences of it.

I'm not borrowing the strict physics. I'm borrowing the vibe — the idea that there's a threshold you cross, after which the relationship between observers stops being symmetric.

The threshold I'm pointing at is the moment, somewhere around October–November 2025, when the publicly accessible AI models crossed roughly the average human IQ. I'll caveat that IQ applied to language models is a rough and contested proxy — these systems are wildly superhuman at some things (recall, breadth, pattern matching across domains) and still sub-human at others (embodied causal reasoning, sustained agency in messy environments). But as a vibes-level anchor for "is the thing in the box, on average, smarter than the person typing into it?" — that line got crossed.

By February 2026, the frontier models that most people can actually use were sitting somewhere around 10–20+ IQ points above the human average. The high-compute variants — the ones you need a budget or an org behind you to really use — were further still.

That sounds like a stat. It's actually a paradigm shift, and I don't think most people have caught up to it yet.

Why the threshold matters

For most of the last few years, the dominant pattern with AI has been using it as a tool. You — the smart human — break the problem down, decide the approach, write the prompt, evaluate the output, stitch the pieces together. The AI is a faster typewriter, a better autocomplete, a tireless intern. You are still the problem solver.

The moment the model in front of you is, on most cognitive dimensions, smarter than you, the optimal pattern flips.

Now the highest-leverage move is to stop being the problem solver and start being the goal definer and orchestrator. You hand the AI the goal, the constraints, the success criteria — and you let it recursively figure out the plan, decompose the work, generate the sub-agents, validate its own outputs, and come back with results.

You're working one altitude up. Sometimes two. Sometimes more.

This is a genuinely hard habit to break for anyone who built their identity, career, or sense of competence on being the one who figures it out. I've been at this for years and I still catch myself diving down into the mechanics of a problem when the right move was to step up a level and let the system handle the mechanics.

The people who have crossed this transition — who actually operate at the altitude the new models are asking them to operate at — are doing something qualitatively different from the people one or two steps behind them. Not faster. Not better at prompting. Different.

And to be really clear — none of this means handing your critical thinking over to the machine. If anything, it's the opposite. The exciting part of working with something smarter than you is that you can get a better answer and ask it to explain why, walk you through the reasoning, teach you the bit you were missing. You can still choose to build understanding, knowledge, and skill — it's just not always required to get the result anymore. That's good exercise for your brain, and it's a much more interesting relationship with a tool than just delegating to it.

The time stretch

Here's the part that's been bugging me the most.

If you arrange people by how close they are to the leading edge of how to think with and work alongside these systems, the gaps between groups are not linear. They're closer to exponential.

Many colleagues and friends would say I'm in the upper range of people working on this stuff — not because I'm special, just because I've made it my full-time job for years, and it overlaps heavily with my personal interests. And from where I sit:

• The 10% ahead of me look genuinely far away. The mental models they're using, the abstractions they're operating at, the pace at which they update — I can see it, I can mostly follow it, but the distance is real and feels like it's growing, not shrinking.
• The 90% behind me look like they're falling further back at an accelerating rate. Not because they're slow, or unintelligent, or unwilling. The terrain is just moving under them faster than the human nervous system tends to update.

It's a stretch. And the stretch is uncomfortable in both directions.

If you're ahead of me, you've probably given up trying to explain certain things to certain people, because the prerequisite mental models aren't there and building them in real-time inside a Slack thread or a meeting is exhausting.

If you're behind me, you've probably had the experience of someone in your network — a colleague, a friend, a boss — talking about AI in a way that sounds like a different language. Not jargon. Something deeper. A whole different way of carving up the world.

And here's what makes it worse: the rate at which the paradigm shifts are happening is itself accelerating. The mental model you built in January is outdated by April. The workflow that was state-of-the-art in March is a beginner pattern by June. The "best practice" article you bookmarked is, by the time you sit down to read it, already wrong about something load-bearing.

The discomfort isn't imaginary. It's structural. The signal volume on the frontier is enormous, and even with AI helping you filter for signal over noise, the rate of paradigm-shifting ideas you'd need to absorb to stay current is faster than most people's bandwidth for absorbing them.

What this does to conversation

The practical consequence I notice most is that talking about ideas across the stretch is getting harder.

A few examples from my own week:

• I try to explain a new way of orchestrating agentic workflows to a peer two altitudes below where the idea actually lives, and we spend the whole conversation building the scaffolding rather than talking about the thing.
• I try to update my boss on something that, to me, is a single coherent insight — and I realise it would take 40 minutes of context to unpack, most of which assumes mental models neither of us has time to build right now.
• I see something on the frontier (a paper, a demo, an offhand comment from someone two altitudes above me) and I genuinely can't tell, in the moment, whether it's a minor tweak or a paradigm shift. I have to sit with it for a few days before I can tell.

This isn't anyone's fault. It's the geometry of where we are.

The temptation, on either side of the stretch, is to retreat — to talk only to people who already share your frame, to dismiss the people behind you as "not getting it", or to dismiss the people ahead of you as "hype merchants". Both of those are easy. Both of them make the stretch worse.

What I think is worth doing

I don't have a clean prescription. This is more of a thought piece than a how-to. But a few things I keep coming back to:

1. Name the stretch when you feel it. A lot of the discomfort in cross-altitude conversations comes from not having a word for what's happening. Calling it out — "I think we're operating at different altitudes on this, let me drop down" or "this is going to take some scaffolding, do you have the patience for it?" — relieves a lot of friction.

2. Build down as often as you build up. If you're ahead of someone, the highest-leverage thing you can do is translate. Not dumb it down — translate. Find the load-bearing idea and give them a version of it they can actually use. The people two altitudes ahead of me who take the time to do this for me are the most valuable relationships I have right now.

3. Be honest about what you don't know. Nobody — and I mean nobody, not the people you see on Twitter, not the lab researchers, not the people writing the long Substack posts — actually knows how this plays out. The honest answer to most questions about the next 12–24 months is "I have a strong intuition and I might be completely wrong". Pretending otherwise is bad faith and it makes the stretch worse.

4. Choose to participate. The unknown is naturally scary, and the rate of change makes it scarier. But the alternative to participating is being acted upon, and being acted upon is worse. Even if you're ten altitudes behind the frontier, the choice to keep showing up and trying to make sense of it is the choice that matters.

The takeaway

I think we're in a strange and historically unusual moment. The thing in the box is, for the first time, generally smarter than the average person using it, and the gap is widening fast. That changes how the best work gets done, who gets to do it, and what it feels like to be on either side of the stretch.

The people behind you on this aren't slow. They're operating in a different frame. The people ahead of you aren't special. They've just been pointing their attention at this for longer.

The bridge between those frames has to be built by the people who can see both sides — and right now, those people are mostly all of us, somewhere in the middle. Pick someone behind you and translate something this week. Pick someone ahead of you and ask them to translate something for you.

Be a little more patient with the people who are struggling with this, including yourself. None of us asked for the terrain to move this fast, and most of us are going to spend the rest of our working lives figuring out how to live on it.

That's the work now. And honestly, I think it's worth showing up for — and an exciting future, if we choose to make it be.