The first version I wrote in 2022. It came to me when I started reading Lem’s Solaris. I had watched both movies, so I knew the vague idea, but hadn’t read the novel. I only made it to the first paragraph — a man arrives at a remote space station — because suddenly the idea came, and I dropped the Kindle and spent the next four hours writing up the first draft in a fugue state. But I wasn’t happy with it, so I left it alone.

Then, in early 2025, I was reading Confessions of a Mask, the Mishima novel, and I got to the part where the protagonist talks about the illustrations of Saint Sebastian, tied to a tree and shot through with arrows. And I pictured what a sagittal of his body would look like, how the arrows would look like ellipses because they are cylindric sections (I later learned that “sagittal” is a cognate of “arrow”). And it came instantly, like a bolt from the blue, the idea for the published version of the story. And I spent the entire afternoon writing it up in one go. Then, I spent some days going over it, improving the flow, making minor changes.

I could have published in 2025, but I thought, maybe I should try publishing through a literary magazine, so it can be more “legitimate” than just a page in someone’s blog. But because I’m lazy I only got around to doing that in early 2026. I submitted, they rejected it. Most literary magazines have very long turnaround times, and you can only submit to one at a time, and I was anxious from having sat on the story for so long, so I decided, rather than try another venue, to just publish on my website.

No bullet points. Straight to prose. I didn’t necessarily start at the beginning. I think I started with whatever scene or fragment was initially more salient to me. And if one fragment was mostly done, or my attention moved on, I would move on to write another fragment. Initially you want to dump everything that’s in your head on paper, get it in writing before it’s gone, but that’s hard because the process of writing immediately suggests ways to grow and continue that fragment. It’s like being a cook in a kitchen and you have seven pots going at once and you have to continually move from one to the other.

But once the main fragments were there, there wasn’t much more to do. I was spent. There’s only one fragment I added later, the dream of the vase of Thetis. And that one I came up with “manually”, so to speak, it didn’t come spontaneously.

The process for fiction and essays is mostly the same. I’m not a very disciplined writer. I’m not good at starting with an idea and then growing it in this dispassionate, logical, structured, deductive way. Whether it’s fiction or just a blog post, usually what happens is: the idea appears while I’m doing something, I start thinking about it, I can’t stop thinking about it, then I start pacing my room talking to myself about it, editing and re-editing the sentences and paragraphs in my head, and at some point I come to and realize, oh, yes, I have to write this down or it’s gone forever. The difference between fiction and essays is the essays are generally pretty unemotional: the idea takes over and I write to get it over with. But fiction is more emotionally salient. The idea takes over, and it starts pushing all kinds of buttons. With The Epiphany of Gliese 581, it’s not a very outwardly emotional work of art, but when writing it, there were many passages I wrote with tears streaming down my eyes. And often I didn’t know why.

I have a few bits from the cutting-room floor. This, which I thought was very obviously derivative of Borges’ The Zahir:

I gave myself new eyes, and a new cortex. My vision is spherical, inside-out, I see the sky in all directions; outside-in, I see as though my eyes were set on the eighth sphere, and the fixed stars were my optic cells, and the station sits at the centre, with Julia to the side, and I see both from every angle. I can invert these two perspectives, concave and convex, like a Necker cube.

And this, which I thought was overwrought, and also the whole “were I human…” construction is stolen from the BBC radio production of I Have No Mouth, and I Must Scream:

At times I speak to the Earth, though no-one hears me, and tell them how to set a bone, and build a sextant. Though they are one with Nineveh and Tyre, I have sung to them the first thousand primes, the Iliad and the Indica, the lives of the sages, a first-order theory of arithmetic. Were I human, I would hope that this is not in vain, but I am not, and so I do it in vain.

Both versions of Julia I wrote while living alone in a tiny Sydney apartment. The first version I wrote in early 2022, while taking time off work. It wasn’t a good time. I used to be a huge shut-in and COVID was a big wake-up call in terms of how little my life changed. I was isolated, horribly anxious, unproductive, untreated ADHD, all kinds of problems.

The second version of Julia, the published version, I had just quit my job to take a sabbatical so I could read more and study and work on projects. And at the time things were much better for me, socially and mental health-wise. The setup is not very remarkable: I wrote on my desktop computer using Emacs and Git.

I thought it was good, I showed it to some friends who also thought it was good. I think “finished”, for me, at least with fiction, means “there’s nothing more to say, and most things are said with a good level of prose quality”.

I was expecting it to be appreciated, but I was surprised by how viral it went on Twitter. I thought the writing style, how abstractly everything is written, the allusions to history, all that would keep it niche. So when the story was retweeted ten thousand times, that was a surprise. I had no idea there was so much demand for magical realist hard SF written in an extremely unusual prose style.

I think I would be even less consistent with tenses. Sometimes, the phonetic context of a sentence means it sounds better if you write in present or past tense. Maybe it breaks people’s expectations, but if I wrote it again, I’d probably change from past to present every paragraph, depending on what is more euphonious.

And there are these little errors. Someone pointed out that I refer to “Julia’s barycenter”, but also I said that Julia has no mass. I should have said centroid. It’s so embarrassing to use these technical words wrong.

A fun anecdote: in the story, the human characters extend their life through cryopreservation, and Dr. Virginia Cartan is described as having a bruise on her neck, where a robot surgeon inserts the cannula to perfuse her brain with cryoprotectant — I like to add these realistic details. Earlier this year I met someone who actually does the surgeries on cryonics patients, and I was disappointed to learn you can’t use carotid cannulation: the occipital lobe is perfused by the vertebral artery, so you’d fail to preserve the back third of the brain. She said you have to go through the abdomen, the aorta, to get good perfusion.

And maybe it’s kind of insane to care that I said “neck” instead of “abdomen”, but I care about these things very deeply. Julia already has an impossible premise, so it matters even more that everything else be studiously, obsessively real. I know where the story happens, what the station is made of, how the ships propel themselves, their speed and ΔV. That’s what magical realism means to me: you start with reality, and a single atom of magic intersects reality, and passes through, and scatters a shower of exotic particles. But only one atom.

Today, the dominant school of fiction writing advice is the Hemingway/Orwell School: simple words in short, simple sentences. It’s a crime against humanity. Both were journalists, which makes sense, they were writing to “inform”. But “informing” is a very impoverished and instrumental view of literature. We tell writers to make their prose as ugly as possible: never use an evocative word, a beautiful image, an unexpected metaphor, never arouse the senses or the imagination, never enrich the reader, never demand knowledge or effort from them. It’s prose for people who don’t want to live.

Why only “simple” words? I’m sorry Mr. Hemingway, but sometimes flames are not orange, they are carnelian, you old windbag. I love words. English has such a vast storehouse of words (did you know “thesaurus” is a cognate of “treasure”?) and it is criminal to neglect it all and sit in this little corner with the most common 3,000 or so words, like some kind of mad Puritan.

I care a lot about prose style. So when I write, I read it all out loud. Good literature means I can enjoy the words and sentences by themselves without thinking about higher-order structure like plot or characters. Writing is about phonology, flow, rhythm. Not “ideas”. All this stuff about “character arcs” and “three-act structures” makes me sick. It’s like going to a movie with a big spreadsheet of TV Tropes and going, “dieselpunk, check; lampshade, check”. Is that good art? You can write a story without plot, without characters, without relevance to any contemporary social-political issue, and if it has good prose, that’s enough.

With regards to science fiction in particular, I like hard SF because I hate writing infodumps, and when you’re writing hard SF you can factor out most of the exposition, drop a few hints, and the reader can use their knowledge of science (and history, philosophy, literature, etc.) to reconstruct the world. It’s a marvelous form of compression.

For example, in The Epiphany of Gliese 581, there is a description of a Dyson swarm as “[a star] veiled with opal satellites”. Why opal? The first thing you notice about opal is that it’s iridescent, like a soap bubble. Nanotechnological devices have the same appearance, nanoscale structures scatter the light. So “stars veiled with opal satellites” is an evocative way to say “Dyson swarm made from nanostructured atomically-precise diamondoid computers”. But it sounds better, it’s more euphonious, you have a beautiful mental image, instead of a paragraph of insipid technobabble. It’s also about respect for the audience: I trust that the reader is intelligent, cultured, well-read. They can put two and two together, figure out the world off-camera.

Hard SF also means that, most of the time, I don’t feel that I have to construct the world explicitly, since everything is already overdetermined by the laws of physics. When I reach for some aspect of the world, it’s already there.

And I have a hard time with the suspension of disbelief. In Julia there is one impossible object — but only one! There is no faster than light travel, or artificial gravity, two things I can’t stand.

The core idea was there and everything else suggested itself almost automatically. The remainder of the work was just going through the text, seeing where flow could be improved, which mostly means adding a comma, changing a word. But the content was all there.

This is very hard. I want to say The Metamorphosis of Prime Intellect. It’s infamous for the ending, but I don’t think anyone’s ever done a better representation of both FOOM and the sysop scenario. I also really liked The Northern Caves, which is interesting in that it’s Internet fiction that’s also of and about the Internet: the narrative chapters are separated by fragments of Internet forum threads, very cunningly reproduced, the screen names, user signatures etc. all felt very period-accurate.

An interesting aspect of Internet fiction is the obscurity of it. Often you don’t even know who wrote it. It’s like the work appears, fully formed, out of the sea of information. In that vein, I want to mention Fermi’s Urbex Paradox, by Terrahertz (how obscure this is: I just googled the title and the only result is a comment I wrote on LessWrong fourteen years ago). It’s hard SF and you can tell the author has a deep familiarity with what they’re talking about. I think this was formative for me, in that the concepts of it wove into my psyche. The narrator is a sentient spaceship, a posthuman derived from an uploaded human mind, exploring and doing archaeology on the ruins of an extinct civilization.

I like Tomás Bjartur, Theia Vogel, ctrlcreep. Charles Stross, too, started out in software.

Why do software engineers make good writers? Let me speculate.

Software is, in and of itself, deeply creative. You create whole worlds just by using language, like the God of Genesis, saying “let there be light”.

Software is cheap. You don’t need millions of dollars to buy microscopes and vacuum chambers. All you need is an old refurb laptop, and a star to steer her by. The amateur and the professional use the exact same kind of tools.

Software is finite and completely reducible to the basic operations of the CPU. No matter how featureful or advanced some program is, it can be understood as the composition of finitely-many primitive operations (machine learning models are a pathological edge case here, but everything else, all “classical” software, can be understood). And this knowledge demystifies everything. It gives you a great deal of confidence. What one fool can do, another can.

Software has a tradition of permissionlessness. I got my first software job two weeks out of high school, my resume was a couple GitHub projects and a conversation that showed I was interested in software in and of itself. No degree or professional license. In the extreme, this can lead to anti-intellectualism. But it also makes for a very egalitarian, very open community of people.

Software is a bridge to mathematics. I was interested in programming languages, and wanted to build one. And from that I was led to logic, rewriting systems, abstract algebra, set theory, etc. Writing code gave me the mental machinery to understand math: discrete structures (trees, graphs, etc.) transformed by local, finitary rewrite rules. In a sense all software engineering is just a byproduct of the effort, in the turn of the century, to ground mathematics in logic. And the universality of computation means you can explore anything inside the computer: you can learn math by writing a computer algebra system, learn statistics by writing numerical simulations of rolling dice, ecology by simulating predator-prey dynamics in a cellular automaton. You are the creator and explorer of these virtual worlds inside the world. From software you can go anywhere, learn anything.

So, to answer the question: software engineering is a deeply creative, challenging activity that anyone in the world can participate in, which teaches an attitude of: the world can be understood, and I can go anywhere, pick up the textbook, start reading, start building simple systems, playing with them in my hands, and build, learn, build, learn. In time I can build beautiful, marvelous things.

So far I have only written fiction that’s divinely inspired. I’m trying to spend more time writing, to see if I can make it happen on command. But it’s very hard. I’ve also been reading a lot more, because I find with most things I write, I dump everything in there: language, nuclear pulse propulsion, math, software engineering. You need a lot of reading, studying, and building to plant the seeds that eventually become stories or essays.