A Rocket for the Mind

A few weeks ago I read that Anthropic's new model, Mythos, had found a 27-year-old vulnerability in OpenBSD.

OpenBSD is the BSD that has spent three decades being the one you cannot break into. Its tagline — only two remote holes in the default install, in a heck of a long time — is the sort of boast an operating system only earns by being unusually careful, for unusually long, in front of unusually motivated attackers. The bug Mythos found would let someone remotely crash any machine running it, simply by connecting. The flaw had survived 27 years of human review, automated testing, and the ongoing attention of one of the most security-fixated communities in computing.

I sat with that for the rest of the afternoon. Then most of the next day.

That is the problem I want to write about.

The awe tax

When I read something like the OpenBSD story, my mind does not go back to work. It hangs. I start sketching consequences — for security, for the software underneath everything I touch, for the kinds of jobs the next generation might do, for the shape of the next ten years. The sketching feels productive. It is not. It is the part of thinking that produces nothing.

I do not think I am unusual in this. Anyone using these tools properly has had the same hour: hands off the keyboard, eyes somewhere past the screen, brain wandering off into what does this mean. The model just did the thing. I should be doing the next thing. Instead I am marvelling.

This is the awe tax. It is the gap between what AI lets me do and what I actually get done, and it widens every time I stop to feel the size of the gap.

The thing under the awe

If I am honest, not all of the hang is awe. Some of it is fear.

Awe and fear feel similar from the inside. Both stop me in my tracks. Both want to be processed before I can move again. The difference is what they pull me toward. Awe wants to admire. Fear wants to retreat. Either way I have stepped off the thing that was carrying me.

When I notice myself pausing to ponder, the useful question is which one it is. If it is awe — fine, give it a minute, then get back to work. If it is fear, that is harder. Fear puts me on defence. People on defence do not build, they brace. They wait to see how bad it gets before committing to anything new. The trouble is that anyone bracing is also losing time at the same rate as everyone else.

I do not want to spend this decade bracing.

The bicycle and the rocket

Steve Jobs used to call the computer a bicycle for the mind. The image came from a study comparing the locomotion efficiency of animals — the condor came out on top, with humans somewhere in the middle, until you put a human on a bicycle. Then the human beat everything. The bicycle did not change what humans could do. It changed how far they could go in a given hour.

I think AI is a rocket for the mind. It is not a metaphor I would reach for casually, but the size of the leap feels closer to that than to another bicycle. A bicycle leaves me in charge of every metre. A rocket asks me to strap in, light it, and stay on board while it does most of the work. The bargain is different. The mistake I keep making is treating a rocket like a bicycle — getting off every few hundred metres to admire the view.

Rockets do not work like that. The point of the rocket is the trip.

Get back on

I am not arguing that nothing about the last few years is worth pondering. The opposite — there is more worth pondering now than at any other point in my life. But the pondering is not the work, and at some point the work has to start again.

So this is a note to myself, as much as anything else. When the next Mythos lands — and there will be one, probably within the month — feel the awe, name the fear, and then get back on the rocket.

The launchpad is impressive. The point was never the launchpad.