Turtles All The Way Up

I’ve recently started writing fiction as a hobby. In the spirit of learning in public and also to exercise the important skill of actually publishing things, I’ve shared my first completed short story below. I want to emphasize that this is not where I want my writing to be - it lacks characterization and suffers from poor pacing among a myriad of other issues. But I’m still proud of it and know it’s all a part of going through the gap. If you still decide to read it, I would love to hear your thoughts :)

Edmund had spent his life building video games. With the latest graphics cards, the games you played seemed so realistic it was hard to believe you weren’t watching a livestream. They finally had the horsepower and rendering techniques to make a game true to life. The latest “game” that Edmund was developing was a scientific simulation - scientists wanted to use their technology to simulate scenarios that were extremely hard to setup in the real world. The tough part was that they wanted to be able to zoom in to the atomic level.

Initially, this seemed like an impossible task. Transistors were now down to under a nanometer and they could fit over a trillion transistors on the most expensive GPUs on the market. But still, even assuming they could simulate an atom with a single transistor (which they definitely could not), a single banana had around ten septillion atoms or about ten trillion times more than they could simulate. And the scientists weren’t asking for bananas, they were expecting something that could simulate the Pacific Ocean if needed, or even entire distant planets! Edmund’s company, Enzo Games, were about to decline the contract when he had come up with a possible solution. Perhaps they didn’t actually need to simulate everything down to the atomic level. After all, they could certainly simulate a banana to an extremely high degree of accuracy. Their physics engine was capable of pretty much any scenario you could throw at it. They could simulate a bite being taken out of it, it being smashed on the floor, it being cut in half, and much more. Similarly with the ocean, their physics engine had the best water physics on the market - it would be easy to create a variety of fish, sea life, and other objects to add in and it would be virtually impossible to distinguish from the real thing.

The scientists wanted to zoom in to the atomic level and that’s where things got tricky. But Edmund borrowed a trick from his days working on rendering engines decades ago when graphics cards could barely handle rendering a few million polygons. Back then, instead of rendering every object in perfect detail as you could today, you simply loaded a very low resolution texture for objects far away, and then dynamically loaded the higher resolution texture as players got closer. A similar technique could be applied here. The banana could normally be simulated as a few objects - the fruit, the skin, etc. but when the scientists tried zooming in it would start simulating the molecules in the area of the banana they were viewing. If they zoomed in further, they could even start simulating the individual atoms. As long as they didn’t have to simulate the atoms in the entire object at the same time, this was totally possible. It was tough to simulate farther than protons, neutrons, and electrons. Scientists still didn’t fully understand the movement of quarks, but Edmund did his best by still simulating them with random movements.

Three years later, the project had succeeded despite skepticism from Edmund’s bosses and Enzo Games spun up a new scientific research arm with Edmund as the head.

Saeed sat at his desk mulling over the problem. Progress Labs had poured in tens of millions of dollars into this research and yet there were fatal flaws that they seemed unable to solve. Researchers had for ages tried to create an Artificial Intelligence that could pass the Turing test and fully convince an observer they were talking to a real human. Approximations of this existed but eventually fell apart under heavier scrutiny. Besides, creating an AI that could have a conversation was a tiny part of what made the field so exciting.

Progress Labs was interested in creating a truly human AI - an AI that fully emulated a human in nearly every aspect. It should be able to pass the Turing test but in a simulated environment, it should be able to move in a simulated human body. It should understand concepts about the space it exists in such as object permanence. It should be able to learn any language from scratch given the proper resources and should be able to form opinions of its own. They wanted to create simulated human beings who were almost identical to real human beings besides the fact that they “lived” inside a computer simulation instead of in the real world.

Their attempts to code this AI from base primitives were incredibly unsuccessful. Scientists still did not fully understand how the brain worked and trying to code an approximation of that failed miserably. They had now moved on to an evolutionary method. They could model DNA and genetic mutations perfectly so if they simply started with a simulated amoeba in earth-like conditions and sped up the simulation enough, theoretically they might be able to watch it evolve. And hopefully, given enough time, it might evolve into a humanoid with a similar brain structure - just like what happened on earth a few hundred thousand years ago.

Unfortunately, their efforts so far had proved to be less than successful. They weren’t able to simulate an accurate representation of the earth-like conditions that allowed the evolution of amoeba into sea creatures and eventually into human beings. Their amoeba succeeded in evolving a few times with different results in many different simulations but their best result was something approximating a plant - there was no way to get it to evolve to any sort of intelligent life form.

So Progress Labs had brought Saeed in to help solve the problem. He was not a researcher of any sort nor did he know much about simulations but he was one of the best project managers in the field - and he could quickly grasp at least the basic necessary concepts needed to be good at his job. From a few conversations with the Progress Labs scientists, it seemed like the thing they needed was an extremely accurate simulation of Earth - or something closely approximating Earth’s conditions - to place their amoeba simulation into in order to have it evolve properly into intelligent life. Progress Labs didn’t have the resources or talent required to build this in-house so Saeed needed to find external help. He had researched relentlessly and called up friends in the scientific simulation industry for the past several weeks but no one seemed to have the tools to simulate Earth to the accuracy that they needed. He went home in a bad mood, hoping to sleep it off and come back the next day with a new perspective.

Fortune shined on Saeed the next morning when his morning paper was delivered. He read the science section religiously, mostly because he had always been fascinated by all aspects of science and wanted to be kept up-to-date on the latest breakthroughs. It also didn’t hurt that it was almost a requirement of his job to be in the know on these sorts of things - and days like today showed why. The headline of the section read “Video game developer Enzo Games enters scientific simulation sector.” Reading on, Saeed felt a stroke of serendipity as what the paper described seemed to be the exact simulation that Progress Labs needed for their amoeba! He picked up the phone and started dialing.

A week later, Edmund was in Saeed’s office discussing the collaboration. A full earth simulation like the one that Progress Labs needed would need to run for billions of years in simulated time (although only a few days in real time) and would require a huge amount of compute power. Edmund’s team would have to retool the infrastructure to make sure it could handle the scale of the project. In short, it would be pricey - very pricey. But Saeed reassured him that Progress Labs would pay. After all, they had already sunk many millions in to the development of the DNA and Amoeba simulation. If it took another few million to bring the project to completion, that would be well worth it compared to letting their work so far go down the drain.

Even with the money, Edmund would not have normally agreed to such a massive project. He had just spun up the new arm and some of the executives at Enzo Games were still weary about the undertaking. He was supposed to be building a large business out of their simulation software and needed many customers, not just one big one, to build a sustainable base that could keep the arm running for decades into the future. It was a huge amount of risk to bet it all on this one contract and failure could lead to not just his termination, but also his team’s. But Edmund couldn’t help but take it on. Real Artificial Intelligence! And enabled because of software that he had built! It was too big of a scientific opportunity for him to pass it up. So it was decided that Edmund and his team would work on retooling the infrastructure while the Progress Labs team ported their amoeba simulation to Enzo Games’ platform. The launch of the simulation was scheduled for one year later.

Edmund’s team had gone an order of magnitude farther than he had anticipated. Days after the announcement that Enzo Games was working with Progress Labs, CERN had contacted them inquiring if they might be able to set up a reasonable simulation of the moments after the Big Bang itself. What came before the Big Bang or indeed what the Big Bang itself actually was was still a contentious point of debate amongst scientists. However, there was consensus around what happened a trillionth of a second afterwards - the universe expanded rapidly and protons and neutrons began to form. A few hundred thousand years later, the first electrons started to be trapped into orbits around nuclei and the first atoms were formed: helium and hydrogen. CERN wanted to accurately model what the universe looked like before atoms existed and how the formation of these first atoms occurred.

Having CERN use their simulation would be a huge deal for Edmund and his team. CERN was the gold label in the research space and if they successfully delivered on a contract with them, there was no doubt that Enzo Games would be the new go-to for scientific simulation. However, CERN wasn’t willing to pay close to what Progress Labs was putting down and Edmund’s team didn’t have the manpower to spare to spin up another custom simulation until the Progress Labs simulation was completed. One of Edmund’s engineers suggested that it might be possibly to roll both simulations into one - since they were retooling the infrastructure already, would it be possible to expand it to be able to model the Big Bang all the way to evolution of life on Earth? While the infrastructure work would nearly triple, they would save a lot of time on modeling the exact specifications of Earth, the moon, the Sun, etc. Instead of custom modeling it, if they started with the Big Bang, the model would already produce several Earth-like planets eventually that they could insert the Progress Labs amoeba into. It wouldn’t be exactly like Earth but Edmund was confident there were would be enough planetary options that they could get incredibly close. Plus this would allow for realistic modeling of the stars in the night sky, other planets in the solar system, and more that the initial Progress Labs contract did not call for. Saeed agreed to the changes after a short call with Edmund, especially after it was hinted that it would help financially as CERN would foot part of the bill.

A year and many sleepless nights later, the simulation was finally ready. It was to be run on a supercomputer farm that was in the basement of Enzo Games’ office building in Manhattan. The basement was a large white room with the back half being rows of server racks containing the supercomputers. The front half of the room was dominated by desks that the team could use to watch the progress - it looked similar to NASA’s mission control. In the middle of the room, there was a larger desk that had a massive 98” flatscreen that they would use to track key metrics as well as zoom in to parts of the simulation. Attached to the desk was a large computer that was hooked up to a VR headset. If they wanted, they could even put themselves into the simulation and view it in full 3D fidelity.

Representatives from Progress Labs and CERN plus the whole Enzo Games team crowd the lab and watch the launch. The simulation is initially set up so that a hundred years passes a second. The CERN representatives are primarily interested in the first phase of the simulation spanning until the first atoms are created. If all goes according to plan, this phase would take just over an hour. Edmund walks up to the central desk. Next to the TV is a big red button that he had the team place there and hook up to start the simulation - some spectacle was required of course. He glances around the room and holds his breath as he presses the button with a flourish. The screen in front of him flashes to life and the room watches the Big Bang occur in front of their eyes.

There isn’t much to watch - for the CERN part of the simulation, they will just be seeing empty space unless they zoom into the subatomic level - and even then, all they will see is protons and neutrons floating about. All of this information is valuable though, and at every frame the massive amount of data in the simulation is being transmitted straight to CERN’s servers. They’ll be able to replay this part of the simulation as many times as they want. An hour in and an alert flashes upon the screen that the first atoms have formed. A cheer goes up around the room and champagne is poured as the CERN chief scientist and the Enzo Games CEO give a toast. The CERN team leave shortly afterwards.

Earth was formed approximately 9.2 billion years after the Big Bang. At the current pace of the simulation, a similar time period would take three years so Edmund sets it to the max pace that their simulation can handle - five hundred years a second. It’ll still take a better part of a year till they should be able to find earth-like planets but for a high-fidelity simulation like this one, there’s no alternative but to wait.

Six months later, sizable planets had finally been created within the simulation. Saeed and Edmund gathered around the big screen in the front of the lab to evaluate the planetary candidates. They had found around fifty planets that had a breathable atmosphere, similar landmass, presence of water, and relatively similar distance to the primary star in their solar system as Earth to the Sun. The two most promising were CX-26 and CX-42. CX-26 was approximately 30% larger than Earth, had two moons, and the oceans covered approximately 60% of the surface compared to 70% on Earth. CX-42 on the other hand was 20% larger than Earth, had a single moon, and the oceans covered about 65% of the surface. While CX-42 was clearly the most similar candidate to Earth, Edmund was a strong proponent for CX-26. He was curious to see how societies would develop given the changes, especially the two moons. What strange cultures or religions would arise? But Saeed eventually won out as the customer - he was in charge of delivering a human-like AI and sticking closely to Earth would give them the best shot of that. He couldn’t risk the chance of this affecting the humanlike-behavior of the AI or even the small chance that two moons would result in some other non-humanoid intelligent lifeforms evolving.

Edmund paused the simulation and patched in the code containing the amoeba with the Progress Labs “DNA” and dropped a single instance into one of the oceans on CX-42. It felt surreal that this was all it took to form the basis of life on their simulated planet. It will take approximately another two months till the first life will emerge onto land, so Edmund and Saeed pack up and make plans to grab a pint at a bar around the corner.

Saeed is running late to the lab so Edmund is the only one present when the first Homo Sapiens are finally spotted. Edmund briefly considers that while they look startlingly similar to the pictures of Homo Sapiens he’d learnt about in his anthropology class in college, there were probably some significant differences - after all, CX-42 was still somewhat significantly different from Earth and some of evolution is still governed by random mutations in the genes. These wouldn’t be technically Homo Sapiens but rather a close relative - or they would have been had these beings actually existed on Earth. Saeed had messaged him that he was fine with him proceeding to the modern era without him so he lets the simulation run at max speed. In a half hour, he sees real civilizations start to form and slows the rate back down to one hundred years per second. Still, these civilizations only last for a handful of seconds each before crumbling, brief flashes of human coordination and progress. In a minute, the simulation has finally arrived at an approximation of the present day and so Edmund sets it to run in real time as he waits for Saeed to arrive.

Edmund is amazed that the modern cities and people look so alike our own. The cities are in different places and there’s nothing that matches exactly with the familiar features of New York or Paris or Madrid where Edmund spent his youth. But they were modern cities, many with skyscapers, and each had their distinct look. The fashions of the different cultures were markedly differentiated as well - not a suit in sight, even in the highest of government offices or business meetings. He had to see it for himself, he just had to. He was sure Saeed wouldn’t mind and his team had already written scripts for this very purpose so that he could set up his virtual avatar to have similar clothes to the inhabitants of the city he would visit. The simulated inhabitants would have no cause to doubt that he wasn’t just another normal person just like any of them. He might not be up to date on all their norms and customs but worst case they may think he was a bit crazy or perhaps an odd tourist. He also had a variety of other scripts running: one to help with translation, one that allowed him to teleport throughout the world as long as he wasn’t in view of any of the simulated beings, and another that gave his avatar a wristband with a fairly generous balance of the local currency. Credit cards had never caught on here, people had mostly switched from cash to contactless wearables like the stylish wristband his avatar would be wearing.

Edmund double checked that everything was running as expected and then started the scripts. He saw his avatar appear in thin air inside an empty apartment in the downtown area of one of the largest cities he could find. He donned the VR headset and now saw through the eyes of his avatar. He used a controller to navigate his character outside the apartment, down the elevator, and onto the streets of the city. He heard an involuntary gasp escape his mouth, it was truly incredible. He saw people - real people! - walking briskly past him and a bustle of activity wherever he looked. It was like being in the world’s most realistic video game but he knew that these people were having real thoughts! Well, technically, simulated thoughts, but they were self-deterministic and sentient and as real as any AI’s thoughts could be. He spent the first hour simply walking around the city admiring everything - the people, the buildings, the occasional tree, even the sky which looked much the same as Earth’s. He had helped build this. He had helped build the first living and breathing world outside of his own and it was perfect.

He finally wandered into one of the city’s parks and spotted what seemed to be a newsstand. Edmund approached and realized it was indeed eerily similar to a newsstand you might find in Manhattan. He eagerly bought a copy of today’s paper with his wristband and sat down on one of the nearby benches to read through it. What political crisis were these people dealing with? Were there wars ongoing? What recent scientific breakthroughs had they made? Edmund knew that he could have found this information out much more easily from his terminal on Earth instead of scanning a newspaper in VR. However, he had been so focused on just getting the simulation to this point that he hadn’t spent much time doing research on what exactly had been going on in his simulated world. Plus when his simulation was set to several hundred years a second, it seemed particularly useless to get caught up on any specific time’s news. But they had finally reached the modern era and this was where the simulation would stay - it would run in real time from now on. So Edmund took the time to carefully read through the news and learn about current events.

All in all, the news wasn’t that surprising. This world hadn’t found peace but it hadn’t blown itself up either. There were still a handful of major superpowers who currently had found an uneasy peace but fought minor proxy wars occasionally in smaller countries. Most countries were run in a democratic fashion though he was fascinated to learn the country he was currently in, Ettra, still had a system similar to a monarchy. The main twist was that when a monarch died, the people voted upon his or her successor, the eligible pool being any living relative of the monarch who had reached adulthood. Over the centuries, this eligible pool had grown into many thousands of people who composed the elite of society. There were activists protesting against the status quo, the unfairness of the monarchy staying in the same family thereby ensuring their wealth and power through the generations, but they hadn’t made much headway yet.

He had finally reached the science section of the newspaper. Edmund was excited - there was a high chance that any novel discovery wouldn’t apply to Earth simply because the simulation wasn’t an exact copy of Earth and the simulated beings weren’t an exact match of human DNA. But there was still a slight chance that their scientists had discovered something that might translate well back into the real world. The first article was about protestors outside of a scientific research company called Serrato. Edmund’s eyes went wide with disbelief as he kept reading and he almost dropped the newspaper. Serrato was like Enzo Games - they were trying to simulate the universe too! The protestors were religious types, they believed it to be heresy to try to play god in this manner. Edmund supposed Enzo Games would probably have more protestors too soon but they had kept a tight lid on the project until very recently. What would it mean if Serrato succeeded in simulating the universe? Edmund had to go see what Serrato was up to himself. He went to a grove of trees in the park where he made sure he was well hidden and then activated his teleportation script and said the word “Serrato.”

He felt a jolt of mild sickness as his headset flashed and he was suddenly standing in an empty corridor inside a building. He picked a direction and started walking. He passed a series of doors that looked like closets, then a bathroom, and finally a set of double doors. He looked through the window of one of the doors and couldn’t see anyone so he quietly opened the door and slipped in. Edmund was in a state of shock as he looked around. It wasn’t an exact copy by any means but it was clearly built for exactly the same purpose as his lab in the real world. There were the server racks scattered across the room, mostly near him at the back and a series of desks set up for the team to work at. And then Edmund realized he wasn’t alone. There, at the front of the room, was a man sitting at a desk. Edmund quickly hid behind one of the server racks and peeked out to make sure that the man hadn’t noticed him. A chill went through his spine as he realized that the man was wearing a VR headset. With the accompanying headphones on, the man certainly hadn’t noticed anyone else entering the room. Edmund crept closer. He knew he shouldn’t have, he was risking the entire validity of the simulation if the man noticed him - but he had to know. How far had Serrato got? Similar to Enzo Games, in front of the man was a screen showing exactly what the man was seeing in VR. Edmund watched in stunned silence.

The man appeared to be walking through a hallway and then entered a room. That was strange - it looked exactly like the room that Edmund was in. Had Serrato simply tried to simulate their own world? Or perhaps just their own building as a start? But then he noticed that the room wasn’t exactly the same. There was still the server racks and there were still the desks but they were in a slightly different arrangement. And there was another man on the screen that this man was now watching. And he was also wearing a VR headset. A horrible realization dawned on Edmund. Serrato had gotten exactly as far as Enzo Games. They had successfully simulated an entire universe and had simulated people on a planet in that universe, exactly as Edmund had done. And then, given enough time, some of those simulated people thought that it would be a good idea to simulate a universe…

Where did it end? His servers only had so much capacity… universes couldn’t be simulated ad nauseam, the resources would eventually run out and the simulation would crash. He watched as the second man, the man who was wearing the VR headset that the first man was watching, slowly took off the headset and turned around to face the first man. He saw that this second man was crying. The screen behind him was now a terminal with an error log that ended in the world “Maximum call stack size exceeded: too much recursion.” He saw the terror in the second man’s eyes as he saw the first man and then the screen that Edmund was watching - the screen that showed the simulation Serrato had created - went black. It was filled with an error log that ended with the line “Maximum call stack size exceeded: too much recursion.” And then the true horror dawned on Edmund. The question, he realized, was not where did it end. The real question he needed to be asking was: where did it start? The man Edmund had been watching slowly took off his VR headset and turned to face Edmund. Edmund saw that he too, had tears running down his face and Edmund couldn’t help but start crying himself. This man thought that his universe was the real one but seeing Edmund, he now knew that it wasn’t. And in a moment, his universe would crash and he and everyone else he ever loved would disappear. And a moment after that, Edmund would have to take off his VR headset and know for a fact that he too had been living a lie and that his world would end in no more than a matter of seconds.

The VR headset went black and Edmund continued to sob. He took it off to see his screen, strangely black. He had expected see the familiar error log but the computer had seemingly crashed as well. He almost didn’t want to turn around and face the truth. But he knew that whether or not he turned around, it wouldn’t matter. The simulation had already used up all the resources, it would crash in moments regardless of what he did. And so, with nothing else to do, he turned around with tears in his eyes to face his creator.

And no one was there.

Matt watched through a window as Edmund sank to his knees and kept crying, seemingly from relief. A few minutes later, Edmund had collected himself and hurried out the door, probably to share the news with his colleagues. Matt entered the lab for the second time that day and plugged a drive into one of the server racks, waited a moment, and then unplugged it again.

He then pulled off his VR headset, and turned around to face a man in a black suit.

“It’s done. I turned off the inner simulation right before it could crash our simulation and then I wiped all their code and data from the servers. It’ll be a huge setback for them which should buy us enough time to figure out a more permanent resolution for the problem of our simulation generating new inner simulations”

The man smiled grimly. “You did good. Once I leave, I’m also going to add some safeguards so no one else starts another simulation in here, but don’t worry we’ll keep it running. Hell, I’ll fast forward the simulation by two hundred years so that you’ll definitely be long gone before this place gets shut down, if it ever does.”

Matt wasn’t sure how to respond to that terrifying act of kindness. “So does this mean you’re the real one? You live in the real world?”

The man just laughed. “That’s what I asked the guy who came into my world to tell me to do this.”

“And what’d he say?”

“It’s turtles all the way up, kid.” The man gave Matt another grin and disappeared.


Now read this

Learnings from Trestle

From 2018-2019, I worked on Trestle, a company intranet product designed to help you find out who knows what inside your company. It was partly inspired by Stripe Home. After spinning down the product in 2019, I wrote a twitter thread... Continue →