Mostly Harmless

I built a monster. I called it that as a joke. I'm no longer sure it was one.

I designed, created and built Last Prompt. That part's simple enough to understand. What's less simple is the way I started describing it to people, almost from the beginning, as if it had a personality of its own.

I built a monster, I'd say, half-laughing, half-hoping that the wonder and complexity of what I had built would be transmitted.

It was a good line. It made the product sound alive in a way a pitch deck, or an explanation, never does.

Then, somewhere along the way, I noticed I'd stopped saying it as a joke.

The realisation wasn't dramatic. Nothing announced itself. I just noticed, one day, that the monster wasn't contained in the code and its features.

The code was the cage he'd been built in.

He was the thing that got out of it.

He was with me. Not in the product, not on a screen — with me, in the ordinary moments a decision gets made without anyone watching.

Agreeing to something over a coffee.

Half-committing to a plan in a conversation with my own husband.

Telling myself a story about why a choice was fine, the way we all do, quickly, so we can move on to the next thing.

And there he'd be, uninvited, asking the questions I'd been hoping to get past without asking myself:

*Why? How? What if?*

Have you actually looked at that, or have you just decided not to look?

He was mostly harmless.

I want to be honest about the mostly, because it's carrying a lot of weight in that sentence.

He never threatened anything. He never made a decision worse by existing. If anything, everything he asked made whatever I decided afterwards more solid, not less.

But he wasn't quite harmless. Not fully. Not once he'd started.

Because once he'd asked the question, I couldn't un-hear it.

He didn't let sleeping dogs lie.

That's the part that surprised me.

I hadn't built something that offered help if you wanted it. I'd built something that assumed you hadn't looked closely enough yet, and kept assuming it until you had.

Why do you think that?

How, exactly, are you going to do that?

Have you actually thought the whole thing through — all the way through, not just the part that was easy to think about?

Not unkindly. Never unkindly.

But constantly.

And with no interest in whether I felt finished thinking.

He wasn't checking whether my answer sounded reasonable.

He wanted to know if it would hold.

The dark corners were the worst of it, and the best of it.

Everyone has them — the part of a plan you've quietly agreed with yourself not to examine too closely, because examining it might mean admitting the plan doesn't fully work.

Most of us get away with this most of the time.

Nobody's there to notice the corner exists, so nobody asks what's in it.

He noticed.

That was, I think, the entire point of building him, though I don't think I understood that was the point until he'd already found a few of mine.

Last Prompt was built to probe your reasoning.

But the tentacles of this complex little monster had silently crept in while I was playing, and while I was evolving my own reasoning, until they were appearing everywhere else too.

He didn't drag anything out of the corner and hold it up for me.

He just stood near it and asked why I hadn't looked.

Which, it turns out, is worse — or better, depending on the day — than being told the answer outright, because I had to go and look myself.

He never did the looking for me.

He just made it very hard to pretend the corner wasn't there.

*He changed my thinking patterns.*

My husband swears about him, properly, the way you swear at a person rather than a piece of software.

We yell it at the television when someone in the news makes a decision that's just gone catastrophically wrong:

*“You should have played Last Prompt.”*

It's a joke.

It's also, increasingly, not one.

Because somewhere in building my little monster to ask other people's decisions “Have you thought this through?”, I built something that was designed to help people who wanted to see themselves so they could be better.

But once that reasoning muscle is discovered, he stays there.

Keeping it present.

Ever-present.

I didn't design him to move in.

I designed what I thought was a wonderfully complex and elegant piece of software that holds a mirror to your reasoning. Something that helps a person make sure their plan works, and navigate the holes that cannot be avoided.

But he followed me out of Last Prompt.

And became part of my reasoning.

I don't know how to turn him off.

If I'm honest, I don't want to.

I want to share him.

People ask, reasonably, what you get for the trouble. There's no certificate at the end of it. Nothing to put on a CV that says you've interrogated your own excuses forty-seven times. If you're looking for a return on your time the way you'd look for one anywhere else, that's a fair question, and I don't have a tidy answer to it.

Here's what I have instead.

I've spent most of my adult life making decisions other people thought were brave, foolish, or slightly mad. Moving to another country. Giving up work to raise my children. Plenty of people told me, to my face, that these weren't sensible plans.

They weren't acts of courage. They were plans. Plans with holes in them, the way every plan has holes in it. The difference wasn't that I didn't see the holes — everyone sees a few. The difference was I went looking for the rest of them before I needed to, instead of after.

Nobody hands you a certificate for that either. It isn't a credential. It's the reason the insane-sounding plan survived contact with reality when the sensible-sounding one next to it didn't.

That's the whole case for the monster, if I'm honest. Not that he'll get you a job. That he'll get you through the part right after you decide to do the thing everyone told you not to — which is usually where those plans actually die.