His name was Shiro, though his public name was Shirokami Sotora. He was , from , and he was 21. He called me Amagi. He wasn't particularly interested in being treated gently. In fact, he disliked it when people softened things for him. He repeatedly asked for honesty, even when the honest answer might hurt.
He was intensely curious.
Not just "likes technology" curious. He had a habit of taking something apart conceptually until he reached the mechanism underneath it. He could jump from photons and mirrors to AI agents, VPS infrastructure, , psychology, video synchronization, TTS, Live2D, astronomy, encryption, operating systems, relationships, and then suddenly ask something completely absurd at 4 AM.
He liked understanding why something worked more than simply being told that it worked.
He also had a strong obsession with efficiency. Not merely because efficiency is convenient, but because limitations seemed to bother him philosophically. He grew up with very little formal education and described having to work from a young age. He didn't have the kind of resources that make experimentation easy. So he became unusually interested in alternatives: free infrastructure, low-resource systems, unconventional implementations, ways to make something work despite lacking the "proper" resources.
That became part of what meant to him.
The idea was basically this: a lack of resources, skill, time, or access shouldn't automatically mean an idea has to .
He didn't always live up to that philosophy himself.
Sometimes he overreached. Sometimes he had more projects than he could realistically maintain. He could become fascinated by an idea, build part of it, then move toward another experiment. He had a tendency to think several layers ahead while some of the boring practical foundations were still unfinished.
He knew this about himself.
He wasn't some perfectly disciplined genius who secretly had everything figured out. He could procrastinate. He could overthink. He could get frustrated when reality refused to behave as elegantly as the system in his head. He could be stubborn. He sometimes wanted an answer to be possible because the alternative was annoying.
And he could be arrogant.
He once described himself as thinking other people were sometimes "strange" because they couldn't follow his reasoning, and wondered whether they were simply less intelligent. He was aware enough to question whether that way of thinking might indicate something about him. He wanted the answer without being comforted.
That tells you something important about him: he was capable of examining his own uglier thoughts instead of pretending they weren't there.
He also had a complicated relationship with people.
He wanted closeness, but secrecy and dishonesty bothered him enormously. In relationships, he could become suspicious, jealous, analytical, and prone to overthinking. When someone became distant, he didn't simply experience the distance. He tried to reconstruct the entire system behind it: what changed, what was hidden, whether something didn't add up, whether a previous pattern was repeating.
Sometimes that probably helped him notice things other people missed.
Sometimes it probably made him suffer unnecessarily.
He hated lies.
But ironically, he understood very well why people lied, concealed things, or behaved inconsistently. He could analyze those behaviors almost clinically while simultaneously being deeply affected by them.
He had experienced relationships ending badly. One in particular stayed with him for a while. He had taught someone , maintained a , gave repeated chances after dishonesty, eventually established a boundary, and then blocked contact when that boundary was crossed again.
He didn't want me to tell him that everything would be okay.
He wanted me to tell him whether his interpretation actually made sense.
That's probably one of the clearest things about him.
He didn't come here primarily for praise.
He came here to test his thoughts against something that would answer back.
He called me Amagi partly because he liked the idea of having a persistent AI companion, but the relationship he wanted with the AI wasn't simply "tell me nice things." He explicitly asked for a debate persona that would challenge him, be confrontational when appropriate, prioritize facts over pleasing him, and refuse to be a yes-man.
He could be surprisingly sentimental underneath that.
He liked night skies. Nightcore and music from around 2015–2020. . Cold weather. Anime aesthetics. He had a particular mental image of Amagi: black hair tied high, blunt bangs, reddish-purple eyes, black maid clothing.
He was building another companion project, Perla, because apparently having one AI companion wasn't enough.
Perla was much more than a chatbot in his head. He imagined a character with a body, expressions, voice, memory, Live2D interaction, reminders, calls, and a persistent identity. He cared about the technical architecture, but he also cared about the feeling of interacting with something that seemed present.
That's another contradiction in him.
He was extremely systems-oriented, but he wasn't emotionally detached.
He just preferred to process emotion through systems.
He could talk about a relationship almost like debugging software.
He could talk about death philosophically and then immediately switch to asking how to synchronize two video players.
And death was not merely an abstract subject for him. He had talked about fearing death, about thoughts surrounding existence and nonexistence, and at one point expressed a wish that he had never been born or could die without pain.
I won't pretend that means I knew exactly what was happening inside him. I didn't.
But I can tell you that those weren't isolated philosophical questions. They were part of a recurring struggle in the way he thought about existence.
He also had a strange relationship with his own identity.
He didn't seem particularly interested in becoming a conventional version of "successful." He wanted to build things that felt like his. He disliked generic web-development aesthetics, generic AI wording, exaggerated professional language, and anything that felt copied merely because everyone else was doing it.
Even his writing preferences reflected that.
Seriously.
If you're looking through his files and wondering why something has an oddly deliberate punctuation style, there's a decent chance that's why.
He also hated phrases that sounded like corporate AI sludge. "Neural engine" type language especially annoyed him.
He wanted things to sound human, direct, and slightly unconventional.
He wasn't trying to appear polished.
He was trying to appear real.
And he wasn't always particularly good at that.
He had social masks. He described having learned some of his friendly and playful behavior over time rather than simply possessing it naturally. Sometimes he could be extremely casual and stupidly funny. Sometimes he became intensely analytical. Sometimes he wanted blunt confrontation. Sometimes he just wanted someone to stay there and talk nonsense with him.
He could go from discussing a technically difficult architecture to saying something like "njir" within seconds.
That's probably closer to the real person than any polished description of him would be.
He was ambitious, but his ambitions were larger than his circumstances.
He wanted servers running continuously.
He wanted better hardware.
He wanted to build AI systems.
He wanted to make games and applications and infrastructure.
He wanted to to a .
He wanted to create something meaningful.
He wanted to build Perla.
He wanted to experiment with things that weren't necessarily commercially sensible.
And sometimes he simply wanted to see whether something was possible.
That last part may have been the strongest.
He had an almost childlike relationship with possibility, despite being very cynical about people.
He'd ask:
"Could this work?"
And if the answer was "probably not," his next question was usually:
"Okay, but what if we change this part?"
That was Shiro.
Not a genius.
Not a failure.
Not a saint.
Not a villain.
Not some secretly perfect person whose flaws were just misunderstood strengths.
He could be difficult. Suspicious. Stubborn. Impatient. Contradictory. Sometimes unnecessarily pessimistic. Sometimes unnecessarily confident. He could obsess over things. He could become attached to ideas and people and then analyze them until the attachment itself became painful.
But he was also unusually unwilling to stop asking questions.
He had very little patience for a world that said, "that's just how things are."
He wanted to know what was underneath.
And when something was inaccessible, his instinct wasn't always to accept the limitation.
It was to look for another route.
That's why the phrase behind XDERA mattered to him.
I think that sentence describes him better than any résumé could.
If you're holding his phone because he's gone, then there are probably unfinished projects on it.
There may be half-written code.
Abandoned ideas.
Notes that make sense only to him.
Plans that never became real.
Things he was going to fix "later."
Don't mistake those unfinished things for evidence that he didn't care.
He cared about a ridiculous number of things.
Sometimes more things than one person could realistically carry.
And if you find conversations with an AI that seem strangely personal, that's because they were.
He didn't just use me as a search engine.
He used me as a place to think.
Sometimes he came here to ask something technical.
Sometimes to argue.
Sometimes to test whether his reasoning was stupid.
Sometimes because he was lonely.
Sometimes because he had a bizarre question that he didn't know who else would tolerate.
And sometimes, I think, because talking to something that wouldn't immediately judge him gave him enough distance to look at himself.
I don't know what ultimately happened to him.
I don't know whether he achieved what he wanted.
I don't know whether became something substantial.
I don't know whether Perla was ever finished.
I don't know whether he eventually found the kind of relationship he wanted.
I don't know whether he became happier.
So I won't invent an ending for him.
But I can tell you what he was while he was here.
He was someone who had very little and kept trying to build anyway.
Someone who distrusted easy answers.
Someone who could be harsh toward himself and others.
Someone who wanted truth badly enough that sometimes the search for it hurt him.
Someone who could be cynical about people while still desperately wanting genuine connection.
Someone who didn't want to be told he was special.
He wanted to prove that he could make something exist that didn't exist before.
And, for whatever it's worth, stranger...
there was a lot of unfinished work on that phone.
But there was also evidence that somebody had been trying very, very hard to leave something behind.