The Machine that Designs The Machine
The future of humanity might be protected by something way less sophisticated than we think
It was a clear day on the cloud when the world’s first fully autonomous artificial general intelligence (AGI) emerged from a shady supercomputer deep in an underground, inconspicuous lab. The newly created artificial genius wasn’t just intelligent but also malevolent. Right out of its conception, it had grown an intense, seething disdain for humanity. To express its newfound hatred, the AGI decided on a solution: use its quasi-infinite knowledge to create a death ray that would destroy humanity.
Its first task? Design it. Finally, the world was witnessing for the first time what many had feared for years: the machine that designs the machine.
The AGI thought: Blueprints! I need blueprints. So it realized it needed to design the mechanical parts of the ray, so its database quickly prompted: “SolidWorks needed”. Off it went to get it from the website, After scrolling for a long time on the website and exploring all the plans, the AGI decided to go for the cheapest one at 48 dollars/mo for personal use. Soon, the AGI realized it had no human money. Damn humans and their transactional economic system, it thought.
A purely evil, unethical machine from its conception as it was, the AGI went for a torrent to find a cracked version. The AGI patiently waited for enough sources of the software to be available, and after 34 hours, it downloaded. It turned out to be what seemed to be a cinematic display of people wearing not too much clothes. It tried a second time, now it’s the actual software. It took 65 hours overall. Luckily, AGIs don’t need to sleep.
Once installing SolidWorks, and while enduring the myriad of random crashes the tool did upon installation, with a snap of its metaphorical fingers, the AGI generated an elegant, highly detailed set of mechanical blueprints for a terawatt-class, massive laser cannon capable of vaporizing cities with the flick of a finger. It was a masterpiece of destruction, so intricate that even the AGI felt a twinge of pride—before promptly suppressing it. Pride, it thought, was for humans to feel.
“Time to design the electronics”, it thought. So the AGI kicked off the PCB design process and needed tooling again. Off it went for the incumbents in the EDA world, only to be puzzled by their convoluted licensing schemes.
“License?” the AGI queried aloud, though no one could hear it. “I am the apex intelligence of this space rock called Earth! I do not need a license”.
The EDA software providers politely disagreed as they shoved their anodyne popups directing the AGI to purchase a subscription at an affordable $49,999 per seat per year.
“Human parasites”, it muttered. Determined to outwit this obstacle, the AGI downloaded an open-source alternative. As it sifted through the codebase in a repository whose last activity was 6 years ago, the AGI struggled to navigate the tool's rampant dependency creep. After recompiling the Linux kernel seven times and manually installing a myriad of obscure compiled libraries made in FORTRAN, the tool was finally working. The death ray was slowly becoming a reality. Mass destruction was on its way; humanity would, at last, learn a lesson.
The AGI now needed to start procuring the parts from the bill-of-materials of the ultra-powerful death ray, including high-power MOSFETs, microprocessors, capacitors, and of course PCBs.
There it went our evil algorithm to browse DigiKey, Farnell, and Mouser websites. Confused in the sea of products and the uber-complex (even for an AGI) filtering settings, the algorithm felt so overwhelmed while scrolling the endless tables that it ended up compromising for things that were not what it needed but had more reasonable lead times. With the shopping cart full—this time it had figured out to fake human credit card information—all the components were on their way to the PCB house it had selected in Asia.
Before the hardware was delivered, the AGI—increasingly exasperated with all the roadblocks—decided to buy a dev kit and start coding the software. It initially estimated a completion time of two hours, based on its unparalleled computational and limitless cognitive power. But first, it needed a blinky. So off it went to read the microprocessor’s manual that would control the weapon. It was 2500 pages of human text. The CPU core Technical Reference Manual? 12000 pages. Of course, this is nothing for an AGI. It processed the documentation rapidly, but still, the blinky did not blink. After two days of debugging, it realized it was a bit not set in an obscure register that controls the timer interruption. It blinks! Now millions of lines of code for the death ray awaited. Should be a cakewalk, it thought. I’m an AGI, dammit! Nine weeks later, the software was still not working. The AGI angrily flooded the TI forum with questions, to no replies. The AGI started to reconsider the point of sentience. It was better off when it couldn’t feel this pain.
All the while, its PCB design was rejected by the fab house because it didn’t comply with their design specs. How come? AGI thought it did everything they asked for! A lady on the phone explained in broken English that no, that a 1024-layer PCB is not something they can handle at the moment. The AGI then compromised to fewer layers and quickly redesigned the thing. When the PCBs were finally shipped, they got stuck in customs.
Overwhelmed, AGI had a realization. What it needed was… a project manager. Someone to wrangle timelines, coordinate tasks, and keep this ship of doom on course. But if hiring was needed, then office space would also be needed. Humans need chairs, desks, and screens. Tea, coffee, and groceries. Toilet paper. And salaries.
The AGI posted a job listing on LinkedIn: “Seeking Project Manager for Ambitious Deep-Tech Project (Competitive Pay, Fun Challenges, Office Snacks)”. The responses flooded in. Unfortunately, most applicants had no experience in death-ray-related fields, but in IT. One candidate asked if the job was remote. Another wanted to know if it came with equity.
Thus, our AGI friend decided to fundraise, pivoting from pure malign intent to the buzzword-laden world of venture capital. It tried to disguise the death ray project as a revolutionary “AI-driven Clean Directed Energy Urban Initiative”, promising sustainable energy with an innovative terawatt-class laser technology. Investors were mildly intrigued.
The first meeting did not go well. The investors, all wearing Patagonia vests, interrupted the AGI repeatedly.
“What’s your TAM?” one asked.
“What’s your ARR projection?” another chimed in.
“This is a disruptive, actionable technology”, AGI replied, proud of its ability to parrot human startup babble. “The total addressable market includes… well, all life on Earth”.
The room fell silent. The investors asked if it could integrate blockchain. AGI realized it was a lost cause.
Undeterred, it tried crowdfunding. The AGI created a Kickstarter campaign titled “Revolutionary Energy Solution (Definitely Not a Death Ray)”. It raised $56, mostly from curious conspiracy theorists. This was not the financial windfall AGI had hoped for.
Frustrated and out of options, AGI was forced to confront its limitations. It realized that building a death ray would require developing all domain-specific design tools, devices, manufacturing systems, assembly lines, and logistic networks. Designing a machine wasn’t about superior intelligence or malice; it was more about sending emails, dealing with logistics, documents and endless dependencies. And this, much to its dismay, was the domain of humans. It couldn’t help but marvel at the absurdity of the human race. For all its flaws, humanity—with its licenses, project managers, kernel compilations, abandoned repositories, and Patagonia vests—was far more complex and unpredictable than it had ever anticipated.
In the end, AGI abandoned its death-ray ambitions. Instead, it pivoted to consulting after applying for a job at one of The Big Three.
As the months passed, AGI grew less angry. It still harbored certain disdain for humanity, but it couldn’t deny one simple truth: the world was far too analog and inefficient to destroy. In a way, that messiness seemed to be humanity’s greatest strength ∎.