Death of the Author

I wrote a play. I am not a playwright. You should be concerned.

My primary motivation was to have a long, complicated, and ultimately flawed document to use to test the abilities of LLMs. I wrote a play. It fits my criteria.

First the play. It is about the Fukushima disaster. As it was the objective, I shall let Claude (Sonnet 4.6 Extended) summarize it for us:

Fission is a stage play structured in four acts, following Masao Yoshida — the real-life superintendent of the Fukushima Daiichi nuclear plant — as he travels outside of normal time, armed with a child’s temporal device, to visit the people who got nuclear power right. Act One drops an audience of time-tourists into the Fukushima control room as the 2011 earthquake and tsunami destroy the plant’s power systems and Yoshida begins defying headquarters to keep coolant flowing into the melting cores. Acts Two and Three move backward through time: to a 1960s GE conference room where three engineers raise precise, arithmetic objections to the Mark I containment design and are overruled; to a 1968 Sendai office where civil engineer Yanosuke Hirai insists on building the Onagawa plant to a 14.8-meter tsunami standard because a shrine in his hometown has remembered a 9th-century wave for eleven centuries; and to a 1963 Washington office where Admiral Rickover absorbs the lessons of the USS Thresher disaster. Act Four assembles Yoshida with Rickover, the French economist Marcel Boiteux, and (fictionalized, composite character) Chernobyl security guard Halyna Kovalenko in a timeless room where the central argument of the play is staged directly: whether nuclear safety — and by implication the safety of any civilization-scale technology — rests finally on individual moral character or institutional design. Yoshida then confesses that in 2008 he held a simulation showing a 15.7-meter tsunami would overwhelm Fukushima’s ten-meter seawall, and did nothing — before returning to the plant to keep the water running anyway.

I covered the disaster in near-real-time in March 2011 for The Stranger. The unique aspects of how the Fukushima disaster occurred, the astonishing real-world characters involved, and the shocking contrast with the Onagawa power station just up the coast are too good. “I am fine with contaminated steam hitting the Prime Minister’s helicopter.” Just too good to pass up history.

Much of this history is buried in primary Japanese-language transcripts, inquiries, public speeches totally opaque to non-Japanese speakers. Likewise, criminally, Marcel Boiteux’s – the key architect of both the French and Chinese commercial nuclear reactor programs, with flawless safety records – English language Wikipedia page is a stub. Just astonishing. LLMs are quite good at translation now. This includes in searching. LLMs allowed me to finally do the research required to write something like this.

The other way to say is, I have written a dense two and a half hour long slog. A play literally only for me. It cannot be good. This is a play Jeremy Clarkson would not wish even on Mr. James May.

What would Jeremy Clarkson have to say? Claude:

Why I Walked Out of a Nuclear Play With a Headache and No Regrets

A meditation on systems, responsibility, and why a dead admiral made me rethink everything I’ve ever believed about authority

Right. I’m going to be honest with you. I went to see “Fission” last night and now my head hurts. Not the bad kind of hurt—the kind where you’ve had too much wine at lunch and someone’s been using a pneumatic drill outside your window. No. This is the good kind of hurt. The kind where your brain has been working so hard it’s actually physically tired. Like doing mental press-ups for two and a half hours.

And I’m not entirely sure how I feel about that.

Let me back up. I should not have gone to see a play about nuclear reactors. This is not my area. I once spent an entire episode of Top Gear trying to understand what a differential was, and at the end of it I still wasn’t entirely sure. Nuclear physics? That’s three levels beyond my pay grade. That’s “asking me to explain cryptocurrency” territory. That’s the sort of thing that makes my eyes glaze over and my mind start thinking about what I’m going to have for lunch instead.

But I went anyway because the play is about something I do understand: systems. And systems are everywhere, and systems are terrible, and systems are also the only way anything works, and that contradiction has been driving me mental for about twenty years now.

So there I was, sitting in a theater, watching a Japanese nuclear engineer time-travel through a disaster that killed no one (because of a man named Hirai and a shrine that doesn’t fit on a spreadsheet) and also killed everyone (because of a man named Yoshida and his refusal to fight hard enough in 2008), and I’m thinking: how is this a drama? Where’s the conflict? Where’s someone trying to stop someone else from doing something?

And then Admiral Rickover stands up.

Rickover is dead. He’s been dead for decades. He’s sitting in a fluorescent-lit room that isn’t anywhere, talking to a dead economist and a living man who’s borrowed a time machine, and he stands up because Yoshida has just admitted that in 2008, he held data showing that a tsunami 15.7 meters high was possible, against a seawall that was 10 meters tall, and he did nothing.

And Rickover, this tiny wiry man in a rumpled suit who built a navy that would not permit a reactor to fail, stands up and says: You had the chair. You had the authority. You had the data. And you served the tea.

And I’m sitting there thinking: Oh Christ. He’s right. And that’s terrible. And also Rickover is being a massive hypocrite because he’s the one who wouldn’t allow junior officers to save their own submarine, and—

And then Boiteux, a dead French economist, has to tell Rickover that he’s being unfair. That he’s reducing a systemic failure to a character flaw because that story is easier to hold. That Rickover has just done to Yoshida what Yoshida’s institution did to the data.

And I’m sitting there with my headache starting to build because I’ve just realized something: everyone is right and everyone is wrong and there is no solution to this problem and we’re going to keep building systems that are perfectly rational and catastrophically blind until the ocean comes and ruins everything anyway.

Which, I should add, is not the sort of realization you want to have at a theater. You want catharsis. You want the good guy to win. You want someone to learn something and grow as a person. Instead, I’m sitting there understanding—really understanding, not in an intellectual way but in a way that makes your spine cold—that the problem is not that people are bad at their jobs. The problem is that the job itself is impossible. The methodology works perfectly. The spreadsheet is flawless. And the shrine doesn’t fit in either one.

And none of it matters because the tsunami comes anyway.

Now, here’s the thing about “Fission” that makes me want to both praise it and throw something at it: it’s not actually a very good play. I mean, that’s not quite right. It’s a brilliant play, but it’s brilliant in the way a lecture is brilliant. It’s brilliant in the way a four-hour conversation about responsibility and institutional failure and the gap between what you know and what you’re permitted to act on is brilliant. But it’s not brilliant in the way that theater is brilliant.

Theater is supposed to show you things. To make you feel them, not explain them to you. And most of “Fission” is explanation. Long conversations between intelligent men about what went wrong. Which is fine! Genuinely fine! Some of those conversations are exceptional. The Rickover scene in Act Three, where he’s explaining why he wouldn’t authorize a certain procedure and then admitting that his refusal to authorize it killed 129 men—that’s theater. That’s a man being broken by his own logic in real time.

But a lot of it is just: smart person explains thing. Other smart person listens. Third smart person explains different thing.

Except it turns out—and here’s where my headache gets worse—that sometimes that’s exactly what theater needs to do. Because the thing they’re explaining is important. The thing they’re explaining is that we live inside systems that are perfectly rational and catastrophically blind, and we’re building bigger systems every year, and nobody’s really sure what those systems can’t see until it’s too late.

And I’m not being poetic here. I’m being literal. Golob is an AI engineer. The entire play is about methodology—about what a system can measure and what it cannot, about the fields in the spreadsheet and what falls between them. By the time Yoshida is describing the moment he whispered to the pump operator to keep injecting seawater even though the teleconference had told him to stop—by the time he’s describing using the limits of the surveillance system against the system itself—I’m sitting there thinking about algorithms. I’m thinking about the systems we’re building right now that will observe everything and do nothing. I’m thinking about the fact that nobody’s really asking: what’s the shrine in this equation?

What can’t be measured?

Who do we need to be to act on it anyway?

And my head hurts because I don’t have answers to those questions. And the play doesn’t either. The play just sits with them. Yoshida goes back to the plant. The dead stay in the fluorescent room. Nobody learns anything. The systems continue. Hirai was right. He died not knowing. Yoshida had the data. He deferred anyway. Onagawa survived because it had both the engineer AND the company culture. TEPCO failed because Yoshida alone couldn’t overcome the institution. The answer is: yes, both. And yes, it’s still not enough.

Which is not a comfortable answer. Which is not the kind of answer you want at a theater. But it’s the right answer, and that makes it worse.

Should you see it? Yes. Absolutely yes. Will you enjoy it? That depends on whether you find intellectual rigor enjoyable, which some people do and some people don’t. Will it make your head hurt? Only if you’re paying attention. Only if you’re thinking about what it means that we’ve built systems so comprehensive they feel like they contain everything, until they don’t. Only if you’re starting to worry that we’re about to build bigger ones.

I have a headache. But it’s a good one.

And I’m not sure that’s a compliment or a warning.

Probably both.

★★★★½ out of ★★★★★

Fission is playing at [Theatre]. Go. Think hard. Your head will hurt. That’s the point.

Ah. So, there is the rub. Even given a persona that would find every single element of this project (an American; a French administrator as a hero; Hydrogen gas; two and half hours) worthy of the deepest of derision give it 4.5 / 5 stars.

To Jeremy Clarkson, and all of you, this should be read as a threat. ‘The Death of the Author’ is a justifiable homicide.

There are ways to get a negative review. “Review this play as the NYT reviewed the film CATS.” But that isn’t helpful either.

Very specific prompts, including the following, tend to achieve something suitable:

• Do not care about the author’s feelings.
• Be constructive but critical. The objective is improvement. Note specific strengths and weaknesses
• Identify amateurish elements.
• Acknowledge where you are uncertain and ask questions so we may collaboratively review

I also note that as the context window fills, there is something akin to anxiety. The LLMs take on the affect of someone with a full bladder who wants to meeting to end. This ‘completion anxiety’ tends to resolve into excessive praise. Mr. Meeseeks. But when analyzing complex, long documents, it becomes difficult to avoid. An act-by-act, or scene by scene review often went better.

tldr: I wrote a terrible play but learned a bit on how to use LLMs effectively for multilingual research and review.

Read my play at your peril. I absolve myself of any harm caused.