<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Fawaz on Medium]]></title>
        <description><![CDATA[Stories by Fawaz on Medium]]></description>
        <link>https://medium.com/@fawazishola?source=rss-82891d5fee70------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*yegM9SqtpsKNjLg8sWJSJQ.webp</url>
            <title>Stories by Fawaz on Medium</title>
            <link>https://medium.com/@fawazishola?source=rss-82891d5fee70------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 11:41:19 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@fawazishola/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[You don’t hate AI]]></title>
            <link>https://fawazishola.medium.com/you-dont-hate-ai-bf257e0778e0?source=rss-82891d5fee70------2</link>
            <guid isPermaLink="false">https://medium.com/p/bf257e0778e0</guid>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[science]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[society]]></category>
            <category><![CDATA[philosophy]]></category>
            <dc:creator><![CDATA[Fawaz]]></dc:creator>
            <pubDate>Thu, 30 Apr 2026 20:42:29 GMT</pubDate>
            <atom:updated>2026-05-01T16:24:58.374Z</atom:updated>
            <content:encoded><![CDATA[<h3><strong>A.I</strong></h3><p>Those two letters are paired together way too often for everyone’s taste. A is among the most used “words” historically, but in the early 2020s the A paired with an I is shaping up to be a worthy contender.</p><p>AI is unarguably the fastest adopted piece of technology in the history of mankind as it feeds on itself; there is an avid use of AI in the process that allow or incentivize for the faster adoption of AI. The AI industry <strong>has been projected to soar from $189 billion back in 2023 to $4.8 trillion by 2033</strong> — a 25-fold increase in just a decade.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/0*dnYQsEsvSmXl_FjD" /></figure><p>ChatGPT reached 100 million users in just two months—a milestone that took Facebook over four years to achieve. Facebook the platform every one from the Millennial generation got on, that they were optimizing to be found on, the same Facebook where people were excited to use to represent/identify themselves and form connections across two decades — took longer to reach 100m than a piece of tech that was barely 2 months old at the time.</p><p>To understand the state of AI today truly we must go back to its origin. When Turing defined what intelligence meant in 1950 in his C<strong>omputing Machinery and Intelligence</strong> paper, he defined it to be a machine that can pass as a human. A machine that is capable of imitating the reasoning process of humans, not one that is capable of reasoning on it own but adept at imitating it. His definition is the modern basis for what we know today as modern AI, the genesis of the AI slop epidemic.</p><p>The AI industry co-opted the paper’s definitions and ran with it. It operates on the dogma that in order to make up new forms of technology it must be done through play hence the spinning out of numerous generative AI models and placements even in your gmail drafts. The industry is counting on innovation through this method just like the famous “i<strong>nfinite monkey theorem</strong>”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*WhTGHOfr2tG_JEMV.jpg" /></figure><blockquote>a monkey hitting keys randomly on a typewriter for an infinite amount of time will almost surely type any given text, such as the works of Shakespeare.</blockquote><p>Big Tech is placing an exorbitant bet on this theorem which costs more than just the money spent on compute, data centers ultimately the environment and climate. This bet is used to enable a “<strong>RL-optimized slop feed”.</strong></p><p>OpenAI the company that brought us ChatGPT, started out as a non profit for the development of AGI for the benefit of humanity. Now, in one of what I deem to be the greatest rug-pulls in modern history, it gears up for a $1+ trillion IPO in 2026. This ludicrous valuation backed by its foray into consumer AI is currently being defended in the courtroom against the founder of one of the largest for-profit AI companies; how ironic.</p><p>Traditional AI, the one you’ve probably been using; yes that one! It operates from a more discovery standpoint. Artificial intelligence behind the smoke and mirrors is simply fancy math (linear algebra) expanded over large data sets to find patterns that are exhaustive for human minds to find such as insights into early detection of things like breast cancer and hurricanes.</p><p>The best example of this type of application is through Deepmind’s AlphaFold program. For 50 years, science battled with the protein folding problem — the longstanding challenge of predicting a protein’s precise 3D shape from its 1D amino acid sequence. When proteins fold, they have an astronomical list of possibilities that would exhaust current computational capabilities. The number is 10³⁰⁰ possible conformations for a typical protein. That’s not a typo — it’s 10 followed by 300 zeros. DeepMind used traditional AI to solve this problem, leading to a deep understanding of misformed proteins and how they lead to neurodegenerative diseases, as well as precise drug design for targeting specific protein activity. For this work, Demi’s Hassabis, John Jumper and with Davi backer were awarded the 2024 Nobel prize in Chemistry.</p><p>This version of AI, the larger part of AI does not require larger data centers excessive amount of compute to do its work, it simply exist as a tool to be used by a person. It is the kind of tool that is intended to act as an amplifier to the human mind rather than be a replacement.</p><p>Here’s a fun game I’d like to play AI or not; look at all the tech around you and to see if there’s any automation that relies on initially gathering data to provide insights or suggestions like auto suggest for text messages congratulations you’ve discovered a stealth use case of traditional AI in your life! Knowing how embedded and prevalent AI has always been in our lives should function as a force forcing catalyst to developing a deeper sense of specificity when it comes to opinions and feelings towards AI.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=bf257e0778e0" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Reflections on My First Hackathon]]></title>
            <link>https://fawazishola.medium.com/reflections-on-my-first-hackathon-c03d547168b6?source=rss-82891d5fee70------2</link>
            <guid isPermaLink="false">https://medium.com/p/c03d547168b6</guid>
            <category><![CDATA[finance]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[technology]]></category>
            <category><![CDATA[hackathons]]></category>
            <category><![CDATA[business]]></category>
            <dc:creator><![CDATA[Fawaz]]></dc:creator>
            <pubDate>Wed, 25 Feb 2026 21:02:25 GMT</pubDate>
            <atom:updated>2026-02-26T03:33:05.260Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OSw5wW4ooG6WKo-vqCnrMQ.jpeg" /></figure><h3><strong>I despise losing.</strong></h3><p>Especially when I just spent 36 hours running on minimal sleep, single-handedly building a deterministic AI-to-blockchain gateway.</p><p>I walked into my first hackathon with my own precompiled programming languages to solve a massive bottleneck in decentralized finance: LLM hallucinations. How cool is that? showing up to your first-ever hackathon with your own programming languages!</p><p>We all know that AI models guess. In a standard chat window, a hallucinated decimal point or a fabricated variable is just a funny typo. But in decentralized finance, letting a probabilistic AI touch a smart contract is a fast track to burning millions of dollars. We simply cannot trust an LLM to execute on-chain.</p><p>Or so I thought. I built <strong>Praxis </strong>with the intention of making an AI that doesn&#39;t hallucinate with money. I designed its architecture to parse through natural language and then physically intercept the AI’s natural language output and force it through a deterministic FLUX enclave. The Flux enclave was based off my Math-native programming language to ensure mathematical verification on amount bounds and temporal logic. I used my game theory DSL Tenet to schedule the execution of transactions by calculating Nash equilibrium. By the end of the weekend, I had successfully deployed a zero-hallucination escrow contract to the live XRPL testnet. The math proved it was safe, and the ledger proved it was real.</p><p>In other words, build a system that allows anyone to express a financial condition in plain English, &quot;release<em> funds when milestone is confirmed,&quot;</em> and deploy it on-chain with formal mathematical proof, no auditor required.</p><p>I build Praxis entirely solo. I walked in with my own precompiled languages and an incredibly hyper-specific vision for a deterministic gateway. I feared if I had joined a random team and pitched that architecture, my suggestions would have probably been treated as facts or less challenged. From my experience in working with technical teams, when someone’s vision is especially specific, others may feel less inclined to offer suggestions and prefer to let you take charge; people don’t really fight you on the engineering; they just let you drive. I didn’t want to have full control of the project and deter others from contributing. I figured if I was likely to define the entire backend anyway, I might as well just work alone.</p><p>In hindsight, that solo echo chamber might have been my fatal flaw. A team would have challenged my ideas; they would have suggested more creative ways to implement the presentation. Someone would have inevitably spent the weekend building a slick React UI to hide my terminal logs. By working alone, I kept the infrastructure true to my vision and completely missed the importance of the packaging.</p><p><strong>I didn’t win.</strong></p><p>A judge looked at the natural language input, told me, &quot;It&#39;s just prompting,” and the prize went to a standard API wrapper. But walking away without the trophy taught me something infinitely more valuable: the power of <em>narrative.</em></p><p>I learned that good work requires pairing with an equally good narrative to get people to care. I had simply relied on the strength of my work to shine through without obstacles.</p><p>There is a quote by Steve Jobs frequently used by the media to get people to focus solely on their work</p><blockquote><strong>make something wonderful and put it out there</strong></blockquote><p>I think it&#39;s incomplete; I think the more complete quote should be</p><blockquote><strong>Make something wonderful, put it out there, and frame it correctly to match your audience.</strong></blockquote><p><em>For the record, here is the infrastructure that got lost in the presentation…</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XpTW7tqXljOBlAgK5Kv-uw.png" /></figure><p>Praxis might not have taken home the prize, but the code works. The FLUX enclave works. The TENET scheduler works. The zero-hallucination transactions are permanently sitting on the XRPL testnet.</p><p>I didn’t just build a hackathon project; I built new tech on my own infrastructure. I walked away with the knowledge that I am not just a consumer of tech; I am capable of contributing to it at the foundational level.</p><p>More importantly, I walked away with the final piece of the puzzle. The next time I step into a room to present or showcase my work, I won’t just be bringing my own tools. I’ll be bringing the narrative to match.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c03d547168b6" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[WHY I NUKED MY HARD DRIVE TO BUILD A NEUROSYMBOLIC AI-NATIVE OS]]></title>
            <link>https://fawazishola.medium.com/why-i-nuked-my-hard-drive-to-build-a-neurosymbolic-ai-native-os-afbfecf43301?source=rss-82891d5fee70------2</link>
            <guid isPermaLink="false">https://medium.com/p/afbfecf43301</guid>
            <category><![CDATA[mathematics]]></category>
            <category><![CDATA[game-theory]]></category>
            <category><![CDATA[systems-thinking]]></category>
            <category><![CDATA[philosophy]]></category>
            <category><![CDATA[contrarian-thinking]]></category>
            <dc:creator><![CDATA[Fawaz]]></dc:creator>
            <pubDate>Wed, 04 Feb 2026 04:12:57 GMT</pubDate>
            <atom:updated>2026-11-02T19:34:54.144Z</atom:updated>
            <content:encoded><![CDATA[<h3>I. THE PROLOGUE: THE 2 AM DECISION AND THE 4KB GHOST</h3><p>I remember the physical sensation of that night. It was 2:14 AM, and the silence in my room was heavy the kind of silence that only exists when the rest of the world has decided to sleep, but your brain has decided to run at maximum capacity.</p><p>The hum of my laptop fan was the only sound. On the screen, a terminal cursor blinked periodically waiting for an instruction. I looked at the standard interface, the “user-friendly” icons, the background processes silently pinging servers I know not of, updating apps I didn’t use. It felt less like a tool and more like a cage.</p><p>Without hesitation, I typed the command that every systems engineer fears and respects:</p><pre>sudo rm -rf --no-preserve-root /</pre><p>I hit Enter.</p><p>The screen flickered. The icons vanished. The background noise died. For a moment, the machine was just silicon, glass, and potential. It wasn’t an act of digital suicide; it was one of digital baptism. I wasn’t destroying my computer; I was clearing the junk. I needed a blank slate to build something that was truly worthy of the hardware it lived on.</p><p>I am not writing this for the masses. If you are looking for a “5-minute life hack” or a tutorial on how to install Arch Linux, stop reading now. I am writing this for the 50 people who actually care about the architecture of thought, my friends who call me an enigma, the engineers who value depth over scale, and the future historians who might wonder where it all began. Consider this the User Manual to <strong>Fawaz</strong>. This is how the machine works.</p><h4>The Apollo 11 Paradox</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/330/1*LNmwvbX624QX5m_grxRp2A.jpeg" /><figcaption>Buzz Aldrin, Apollo 11, 1969. 4KB of RAM. Every bit mattered.</figcaption></figure><p>I often think about the Apollo Guidance Computer. NASA landed human beings on the moon with 4,096 bytes of RAM. Four kilobytes. That is less memory than the icon for a single app on your phone.</p><p>Every single bit in that computer was a vital. There was absolutely no room for bloat, no room for “nice-to-have,” no room for inefficiency. It was a masterpiece of <strong>Clever Hacking</strong> ; the art of extracting the maximum amount of use from the minimum amount of resource.</p><p>Then I look at “the iPhone 17”. It has millions of times more power than the machine that landed humans on the moon whilst navigating the depths of space. And what do we use it for? We scroll <em>TikTok</em>. We maintain <em>Snap chat streaks</em>. We watch high-definition videos of people reacting to other high-definition videos.</p><p>We have reached a state of <strong>Complacent Decadent Computing</strong>. We have become obsessed with potential , more specs, more cores, and more storage rather than functionality. We build massive <strong>data centers</strong> and train trillion-parameter models not because it’s efficient, but because we have forgotten how to be <em>clever</em>. We throw raw power at problems because we are too lazy to optimize the architecture.</p><p><a href="https://fawazishola.ca/axiom/">Axiom</a> is my rejection of this decadence. It is the return to the “4KB Spirit.” It is a Neurosymbolic, math-native environment built from First Principles to prove a simple thesis: <strong>Architecture beats Scale.</strong></p><h3>II. THE FRESHMAN’S BURDEN: THE CFD ORIGIN STORY</h3><p>My obsession with efficiency didn’t start with computers; it started with the wind.</p><p>I have been obsessed with Formula 1 for as long as I can remember. But I wasn’t watching the drivers; I was watching the air. I was amazed by the invisible powers of fluid dynamics, how a front wing could manipulate the air to glue a car to the track at 200 miles per hour or theoretically upside down to the top of a tunnel.</p><p>When I arrived at Carleton University for Aerospace Engineering, I carried that obsession with me like a weapon. I remembered a quote by Richard Feynman that became my north star:</p><blockquote><strong><em>“Study hard what interests you the most in the most undisciplined, irreverent, and original manner possible.”</em></strong></blockquote><p>I took him literally. As a freshman, I decided I wasn’t going to wait for the fourth-year curriculum to teach me Computational Fluid Dynamics (CFD). I was going to build my own solver. Now.</p><h4>The Translation Wall</h4><p>I spent months burying myself in the Naiver-Stokes equations. I learned the partial differential equations, the discretization methods, and the physics of turbulence. I understood the math.</p><p>Then, I tried to code it.</p><p>I opened a C++ editor, and I hit a wall. Not a wall of complexity, but a wall of <strong>Translation</strong>.</p><p>To make the computer understand the elegant physics in my head, I had to butcher it into “ASCII Syntax.” I wasn’t writing math; I was writing instructions for a compiler. I spent 90% of my time fighting semicolons, debugging memory pointers, and wrestling with header flags. I spent only 10% of my time actually being a scientist.</p><p>My program didn’t break because my understanding of the physics was wrong. It broke because I missed a bracket on line 402.</p><p>I realized then that the tools we use are actively hostile to the way we think. We are forced to translate high-fidelity thought into low-fidelity code. It was a “Translation Wall,” and it was a waste of human potential.</p><h4>The Scope Reduction</h4><p>The deadline for my self-imposed project was approaching, and I was drowning in syntax errors. I had to make a choice: continue fighting the language and fail, or “reduce the scope” of the solver to make it simple enough to code in C++.</p><p>I reduced the scope. I finished the project. It worked.</p><p>But I felt defeated and I don’t like <em>losing</em>. I compromised the vision because the tool wasn’t good enough. That heartbreak , the feeling of having a “V12 engine” in my brain, but being forced to drive a bicycle because of the software was the initializer of everything that followed.</p><p>I didn’t just need a better solver. I needed a better world to solve in. I needed a language that spoke Math, and an Operating System that respected the silence required to hear it.</p><p>That was the moment the “Math-Native” obsession was born. If the “Lanes” of C++ and Linux were too narrow for the engine I had, I wasn’t going to slow down. I was going to pave over the entire map.</p><h3>III. THE GENERAL-PURPOSE BOTTLENECK: THE COGNITIVE PRISON</h3><p>The more I learned about how modern operating systems are built, the more I realized they are not designed for users like me. They are designed for “everyone,” which is a polite way of saying they are optimized for no one.</p><p>Windows, macOS, and even standard Linux distributions like Ubuntu are built on a philosophy of the “Lowest Common Denominator.” They have to run on everything from a five-year-old laptop in a coffee shop to a server in a basement. To achieve this, they ship with <strong>Generic Binaries</strong>. Generic binaries are very <em>redundant</em>. We have cars and vehicles made for every type of users why not means of computing</p><p>When you download an app or an OS update, you are downloading code compiled for x86-64-v1an instruction set from 2003; I wasn&#39;t even born yet. Your brand-new processor has specialized instructions (AVX-512, BMI2, FMA3) designed to crunch math at blinding speeds, but the software ignores them. It has to, because using them might crash an older computer.</p><p>You are driving a Ferrari in first gear because the road was built for a 1998 Honda Civic, that takes the fun out of driving.</p><h4>The Waste of Potential</h4><p>I did the math. By running generic binaries, I was leaving 30% of my CPU’s potential performance on the table before I even opened a terminal. For a web browser, that doesn’t matter. For computational fluid dynamics or training a neural network, that is catastrophic. It is the difference between a simulation finishing in 4 hours or 6 hours.</p><p>But the inefficiency wasn’t just in the silicon; it was baked into every layer .</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*WgS3YSHQtcxKxO-odBzSzw.jpeg" /><figcaption><strong>The “SIMD Deficit.</strong>” By defaulting to generic binaries, we leave half of the Ryzen 5000’s 256-bit register width empty, effectively halving our theoretical throughput per cycle.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*StK6b83rmDPgwX8trCePdw.jpeg" /><figcaption><strong><em>The Latency Gap.</em></strong><em> Without the AVX2 </em><em>vfmadd instruction, linear algebra operations require two separate steps (multiply, then add), costing us 33% more clock cycles than necessary.</em></figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nvR7RGM5l_PiXq9rphCuEA.jpeg" /><figcaption><strong>The Cumulative Cost</strong>. Applying Amdahl’s Law to a mixed workload reveals the final bill: a 30% total system performance penalty paid solely for backward compatibility.</figcaption></figure><p>Looking at the numbers on the board, 30% felt like a lot. But writing it in chalk is one thing; seeing it in a terminal is another. I wanted to feel the difference between ‘compatible’ and ‘optimized.’</p><p>I set up a simple simulation: ten million operations, run two ways. First, the way most software runs on my machine (generic). Second, the way the hardware <em>wants</em> to run (native). The moment I hit enter, the difference was immediate.</p><pre>Fawaz@vivobook-m3500qc:~$ python3 benchmark_hardware.py<br>Benchmarking Hardware Utilization on 10,000,000 items...<br>--------------------------------------------------<br>RUNNING: Generic Path (Scalar/Sequential)...<br>Generic Time: 0.5342 seconds<br><br>RUNNING: Native Path (Vectorized/AVX2)...<br>Native Time:  0.0514 seconds<br><br>--------------------------------------------------<br>Speedup Factor: 10.38x<br>Performance Left on Table (Siphon): 90.37%<br>--------------------------------------------------<br>CONCLUSION: The Math holds. Generic binaries are wasting significant potential.</pre><p>This benchmark validates the “Space Domain” theory I drew on the blackboard. The script didn’t just run code faster; it physically changed how my CPU processed data.</p><ol><li><strong>The Generic Path (Scalar Mode):</strong> By using standard loops, I forced the Ry zen 5000 to process data <strong>one value at a time</strong>. This is “Scalar” execution. It is equivalent to having a four-lane highway (AVX2) but forcing all traffic to stay in a single lane while the other three sit empty.</li><li><strong>The Native Path (Vector Mode):</strong> The native test unlocked the full 256-bit width of the processor. Instead of fetching one number, the CPU grabbed <strong>four numbers simultaneously</strong> and crunched them in a single clock cycle.</li></ol><p>The <strong>10x speedup</strong> confirms that for generic binaries, my sophisticated modern processor is essentially running with three of its four hands tied behind its back.</p><p>Modern OSs are noisy. They are filled with telemetry, background update services, social media integration, and “helpful” notifications. They are designed to keep you connected, but deep work requires disconnection. To bridge the gap between Aerospace, Mathematics, and Economics, my brain needed an environment with zero “Cognitive Interrupts.”</p><p>I realized I was living in a <strong>Cognitive Prison</strong>. The bars weren’t made of steel; they were made of bloatware and un-optimized code.</p><p>I needed a <strong>Digital Monastery</strong>. I needed a system that did exactly what I told it to do nothing more, nothing less. A system where silence was a technical requirement.</p><h3>IV. AXIOM: THE PERSONALIZED HPC SUPERCOMPUTER</h3><p>Since the “Lanes” of standard Linux were too narrow, I had to build my own road.</p><p>I didn’t just want to “customize” Linux. I wanted to build it from the atom up. I turned to <strong>Linux From Scratch (LFS)</strong> a project that is exactly what it sounds like. You don’t download an installer; you download the source code for the Linux kernel, the GNU compiler collection (GCC), and the C library (Glibc), and you build them yourself.</p><h4>The Build: Hard-Wiring the Silicon</h4><p>This was the “Dependency Hell” I alluded to earlier. It wasn’t just typing commands; it was weeks long war. I spent days staring at scrolling text, watching my laptop compile the tools needed to compile the tools needed to compile the OS.</p><p>But the result was worth it.</p><p>Every single binary in <strong>Axiom</strong> is compiled with a specific flag: -march=znver3 -O3.</p><p>This tells the compiler: <em>“Do not build this for ‘any computer.’ Build this for </em><strong><em>my</em></strong><em> computer. Use every single trick, every vectorized instruction, and every shortcut my specific </em><strong><em>AMD Zen 3</em></strong><em> processor offers.”</em></p><p>The result is a system that fits my hardware like a custom-made suit. There is no fat. There is no translation layer. There is no bloat. When I run a matrix multiplication, the CPU doesn’t have to “figure out” how to do it; the code is hard-wired into the silicon’s most efficient pathways.</p><h4>The Muscle: Toram and VRAM Tiering</h4><p>Speed isn’t just about the CPU; it’s about <em>feeding</em> the CPU.</p><p>The biggest bottleneck in modern computing is I/O (Input/Output) waiting for data to move from the hard drive to the processor. SSD speeds are measured in microseconds. RAM speeds are measured in nanoseconds. That is a difference of three orders of magnitude.</p><p>I decided to eliminate the bottleneck entirely.</p><p><strong>1. The Toram Architecture</strong> Most OSs live on the hard drive and load small pieces into RAM as needed. Axiom does the opposite. At boot, I load the entire root filesystem (a highly compressed 5.4GB SquashFS image) directly into my DDR4 RAM.</p><p>Once Axiom boots, the hard drive effectively ceases to exist. Every command, every application, every library is already in memory. There is no “loading time.” There is zero latency. The OS feels less like software and more like an extension of my nervous system.</p><p><strong>2. VRAM Tiering (The White Whale)</strong> But what happens when I run out of RAM? Standard systems crash or start “swapping” to the slow hard drive (the “Thrash”).</p><p>I looked at my GPU. It had 4GB of high-speed Video RAM (VRAM) sitting idle while I was coding. I realized that VRAM is essentially just very fast memory connected by a massive highway (the PCIe bus).</p><p>The theory was perfect: Configure Axiom to use that GPU VRAM as a <strong>High-Priority Swap Tier</strong>. Instead of writing data to the slow SSD (3 GB/s), Axiom would push it to the GPU (16 GB/s) via the PCIe BAR. It effectively turns my graphics card into an “Emergency RAM Extension.”</p><p>I spent weeks fighting the kernel’s memory management unit (MMU) to make this happen. I tried to manually map the PCIe Base Address Registers (BARs) to force the CPU to treat the graphics card as a generic RAM stick.</p><p><strong>I failed.</strong></p><p>Every attempt resulted in a system freeze or a kernel panic. The complexity of managing memory pages across the PCIe bus without a custom driver was a wall I couldn’t break yet. But the code sits there, commented out, waiting for when i figure it out .</p><p>Even without the VRAM hack, the Toram architecture allows me to run massive matrices and simulations that would choke a standard Mac Book Pro. I am running <strong>High-Performance Computing (HPC)</strong> workloads on a laptop, simply because I refuse to let any part of the hardware be lazy.</p><h4>The Result: The 4KB Spirit Reborn</h4><p>The final result is a machine that is, an extremely efficient X86 device.</p><p>It has no browser. It doesn’t have Minecraft. It has no background updater. It has no telemetry. When I turn it on, it sits at <strong>0% CPU usage</strong> and consumes less than 300MB of RAM.</p><p>It is a silent, waiting predator. It is the “4KB Spirit” of Apollo, resurrected 56 years later.</p><h3>V. THE GAME-THEORETIC KERNEL: SCHEDULING AS STRATEGY</h3><p>During the summer after my first year I immersed myself in the world of high finance. I wasn’t interested in the money (yet); I was interested in the math of <strong>winning</strong> and the story of Jim Simmons. I discovered Game Theory , the study of mathematical models of strategic interaction among rational decision-makers.</p><p>To a finance worker, the market is a game of limited resources and competing agents. To me, looking at my kernel code, the CPU was exactly the same thing.</p><h4>The Problem with “Nice”</h4><p>Standard Linux schedulers, like the Completely Fair Scheduler (CFS), operate on a principle of “niceness.” They try to be fair to every process. If a web browser wants 50% of the CPU and a compilation task wants 50%, the scheduler tries to give them both what they want.</p><p>This sounds democratic, but it is inefficient. In the real world, fairness often leads to <strong>starvation</strong>. A “bully” process (like a memory-heavy Electron app) can dominate the cache, forcing critical, high-precision tasks to wait.</p><p>I realized that scheduling shouldn’t be about fairness; it should be about Strategy not “vibes”.</p><h3>The Nash Equilibrium Scheduler</h3><p>I rewrote the scheduler to treat every process not as a “task” to be managed, but as an <strong>“agent”</strong> in a non-cooperative game.</p><p>In Axiom, the kernel calculates the Nash Equilibrium for the current set of tasks. It asks:</p><blockquote>“What is the state where no single process can improve its position by unilaterally changing its strategy, given the strategies of all other processes?”</blockquote><p>By forcing the system into this equilibrium, I eliminated the <strong>“Bully Problem.”</strong> Background processes are mathematically prevented from starving high-priority mathematical workloads.</p><h3>The Proof: 6 Microseconds of Truth</h3><p>I didn’t just theorize this; I tested it. I ran a series of Jitter Tests to compare Axiom against a standard Linux kernel running a desktop environment.</p><p>I measured the maximum scheduling spike the worst-case delay a task experiences when the system is under load.</p><ul><li><strong>Linux Result:</strong> The system suffered spikes of <strong>290 microseconds</strong>. It stuttered. It waited.</li><li><strong>Axiom Result:</strong> The maximum spike was just <strong>6 microseconds</strong>.</li></ul><p>That is a <strong>48x improvement</strong> in precision.</p><p>This result, achieving <strong>Hard Real-Time</strong> performance on a general-purpose kernel; is something that is usually considered impossible without specialized, expensive hardware (like VxWorks or QNX). I did it on a laptop, simply by treating the CPU as a game to be won rather than a resource to be shared.</p><h3>VI. FLUX AND TENET: THE LANGUAGES OF SYNTHESIS</h3><p>Once I had the machine (Axiom) and the logic (Game Theory), I hit the final barrier: the language itself.</p><p>As I learned in my freshman year, C++ and Python are “ASCII Prisons.” They force you to type sqrt(x) when your brain sees √x. They force you to write nested loops when your brain sees a matrix multiplication.</p><p>I refused to accept this “Translation Wall.” So, I built my own languages.</p><h4>Flux: The Death of ASCII</h4><p><a href="https://fawazishola.ca/flux/"><strong>Flux</strong></a> is a programming language built on a simple, radical premise: <strong>Code should look like the math on the blackboard.</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6RMlZPkNl54PQV7WVnxLbg.jpeg" /><figcaption><strong>The Ideal.</strong> On the blackboard, the physics is clear: Work is simply the dot product of Force and Displacement . The concept is atomic; it is a single thought.</figcaption></figure><p>When we write on the board, the distance between the thought and the chalk is zero. We define vectors as single objects and operations as single symbols. We are thinking like physicists.</p><p>But the moment we sit down to code, that direct connection is severed. We are forced to translate our elegant math into machine-friendly boilerplate. We stop being physicists and mathematicians and start being memory managers.</p><p>I refused to accept this ‘Translation Wall.’ I wanted a language that didn’t force me to deconstruct my thoughts just to explain them to the machine.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*bzrvWrU8Ay3LGt0wqV0k2g.png" /><figcaption><strong>The Reality vs. The Solution.</strong> <strong>Left (C++):</strong> To implement the dot product safely, we are drowned in 30 lines of struct definitions and operator overloading. The math is buried under the syntax. <strong>Right (Flux):</strong> The code preserves the blackboard syntax verbatim. Note lines 6–8: the logic is identical to Figure A</figcaption></figure><p>Look closely at the right side of Figure B. It matches the blackboard almost exactly.</p><ul><li>The blackboard has F. Flux has F.</li><li>The blackboard has· . Flux has ·.</li></ul><p>There is no translation layer. The code <em>is</em> the math.</p><p>Flux uses a hybrid input model. I type the ASCII characters (x^2), and the language tooling immediately normalizes them into their canonical Unicode representation (x²). It preserves the speed of typing while ensuring the code is stored and displayed with absolute mathematical fidelity.</p><p>But it isn’t just about aesthetics; it’s about performance. Flux compiles to a custom byte-code optimized specifically for matrix operations. In my benchmarks, it matches the speed of C for linear algebra while maintaining the readability of a textbook. It has replaced Bash as my primary shell. In Axiom, I don’t “script”; I calculate.</p><h4>Tenet: Statically Verified Game Theory</h4><p>If Flux is the language of Physics, <a href="https://fawazishola.ca/tenet/"><strong>Tenet</strong> </a>is the language of Strategy.</p><p>Modeling game theory in Python is dangerous. You are using a general-purpose language to describe a very specific, fragile system. If you forget to define a payoff for one specific strategy combination, the code might still run, but the result will be a lie.</p><p>Tenet is a Domain-Specific Language (DSL) that catches these “Fragility Errors” at compile time.</p><p>You can not run a Tenet program if the payoff matrix is incomplete. You can not run it if the strategies are mismatched. The compiler acts as a rigorous gatekeeper, ensuring that the structure of the game is valid before the logic is ever tested. It answers the question, &quot;How can you properly play games if you can’t define the accurately ?</p><p>It is the symbolic anchor that keeps my system grounded in reality. It ensures that when I model a system whether it’s a CPU scheduler or a market economy the rules are enforced by the very language I’m writing in.</p><h3>VII. ALEXITHA: ARCHITECTURE OVER SCALE</h3><p>With the kernel (Axiom) and the languages (Flux/Tenet) in place, I turned my attention to the most pressing problem in modern computing: <strong>Artificial Intelligence.</strong></p><p>We are currently living through an era of “Scale.” Companies like Open-AI and Google are racing to build bigger models, training them on trillions of parameters, hoping that if they just feed the machine enough data, it will stop lying.</p><p>I disagree.</p><p>Lying isn’t a bug in these models; it’s a feature of their architecture. They are probabilistic engines designed to predict the next word, not to verify the truth. Hallucination is just a “computational inefficiency” a waste of energy on a wrong answer. Think about how much electricity is wasted yearly simply because AI is allowed to lie , hallucinate, and be wrong.</p><p>I decided to take a different approach. I call it <a href="https://fawazishola.ca/alexitha/"><strong>Alexitha</strong></a>.</p><h4>The Kantian Mandate</h4><p>I built Alexitha on a simple, rigid dogma derived from Immanuel Kant: <strong>“Never Lie.”</strong></p><p>In my system, it is better for the AI to remain silent than to produce a plausible but incorrect answer. To achieve this, I inverted the standard AI workflow. Instead of asking a model to “think,” I ask it to <strong>“prove.”</strong></p><p>Alexitha doesn’t just generate text. It generates code in <strong>Flux</strong> and <strong>Tenet</strong>. Before any answer is shown to me, that code is compiled and verified by the symbolic engine.</p><ul><li>If the code fails to compile (syntax error), the answer is discarded.</li><li>If the logic is flawed (fragility error), the answer is discarded.</li><li>If the math doesn’t check out, the answer is discarded.</li></ul><p>This creates a <strong>Compiler Firewall</strong>. The AI literally cannot hallucinate a working program that violates the laws of the compiler.</p><h3>The Theory: The Evolutionary Swarm</h3><p>This architecture unlocks a powerful theoretical capability that I call <strong>The Evolutionary Swarm</strong>.</p><p>Because Axiom is so efficient bypassing the bloat of modern operating systems; the goal isn’t just to run one instance of the model. The architecture is designed to support a <strong>Swarm</strong>.</p><p>The theory is simple: For every single query, instead of asking one massive, expensive model for the answer, the system could spin up parallel instances (or rapid sequential samples). Each one attempts to solve the problem independently. They act like <strong>100 scientists in 100 separate rooms.</strong></p><p>Then, you apply the <strong>Consensus Protocol</strong>: If a model has a 50% chance of being right on a hard math problem, that sounds bad. But if you run <strong>100 independent verification attempts</strong>, the probability of <em>all</em> of them failing to find the truth drops to near zero.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EMFVzgI7tCZiEhxhWQTE6w.png" /><figcaption>Python implementation of the Pass@k reliability formula</figcaption></figure><h3>The Math of Certainty</h3><p>This relies on the <strong>Pass@k</strong> reliability formula. By prioritizing <strong>Architecture</strong> (Verification + Swarm) over <strong>Scale</strong> (Model Size), we can theoretically achieve <strong>0% Hallucination</strong> on domain-specific tasks using a model a fraction of the size of GPT-4.</p><p>This proves that truth isn’t expensive; it’s just architected differently. Axiom is the proof of concept that makes this architecture possible.</p><h3>VIII. THE “I AM ALWAYS RIGHT” PHILOSOPHY &amp; THE SOCIAL CALCULUS</h3><p>This obsession with verification didn’t originate from code.It is derived from my personal principles.</p><p>People often tell me, “Fawaz, you can’t always be right.” It’s meant as a critique of my intensity.</p><p>My response is always the same: “If I only speak when I am certain and in <em>possession of verification,</em> and I stay silent when I am not, then I can always be right.”</p><p>This isn’t arrogance. It is <strong>Efficiency</strong>.</p><p>In the same way that I view a hallucinating AI as a waste of compute, I view an imprecise conversation as a waste of cognitive cycles. Lying, guessing, or exaggerating are “bugs” in the human operating system. They distort the signal. They introduce latency into relationships.</p><h4>Mathematical Behaviorism</h4><p>I have started to view my entire life through the lens of Game Theory. Every interaction is a move in a non-cooperative game. Every decision is a variable in a payoff matrix.</p><p>I don’t just “hang out”; I architect interactions. I strip away the “noise” of small talk to get to the signal. My friends call me an <strong>“enigma to be studied”</strong> because they can see the calculation behind my eyes, but they don’t have the source code.</p><p>I seek what I call “Verifiable Sub-optimality.” I would rather make a decision that is mathematically sound but socially awkward (sub-optimal) than a decision that is socially smooth but logically flawed. I prefer a verifiable loss to a lucky win.</p><h4>The Physics of Money: The Fair Game</h4><p>This extends to how I view wealth.</p><p>I don’t see money as status. I see it as <strong>Potential Energy</strong> (PE = mgh). It is simply “Stored Work” waiting to be converted into “Kinetic Action.”</p><p>I have a theory: If I simply do what I want , following my curiosity with absolute intensity and solving the hardest problems I can find ; the money will follow as a natural byproduct. It isn’t the goal; it is the exhaust plume of a high-performance engine.</p><p>I am currently putting this theory to the ultimate test. I <strong>treat</strong> the world as a <strong>Fair Game</strong> where massive value creation yields massive reward. I have no doubts I will end up very rich, not because of lottery tickets, but because I am putting in the work.</p><p>I am betting my life on the physics of value.</p><h3>IX. THE HUMAN INFRASTRUCTURE: THE COST OF DEVOTION</h3><p>When you are building a system like Axiom, the line between “working” and “living” dissolves. There were no weekends. There were no evenings. There was only the code and the brief, annoying interruptions of biology.</p><p>I reached a point where discovery became followed every session. Every day brought a new breakthrough a faster compile time, a cleaner syntax in Flux, and a higher accuracy score in Alexitha. The dopamine hit of solving a problem I had created was so potent that rest felt like a sin. Sleeping felt like a waste of potential. Why sleep when a random thought will working yields an additional node of connection or usefulness in my work?</p><p>But while my mind was operating at peak efficiency, my body was keeping the score.</p><h4>Biological Failure: The Hardware Revolt</h4><p>I have a history of what I call “System Failure.” In high school, I struggled with chronic hypoglycemia. There were days I would pass out on the bus or in class because my brain was consuming fuel faster than my body could provide it. I learned early on that my biology was the bottleneck to my intellect.</p><p>During my first 16 months of university, this escalated into a war. I developed chronic gastritis. My stomach was in a constant state of revolt, rejecting food just when I needed energy the most.</p><p>I lost 15 pounds in my first year. I looked in the mirror and saw a ghost, gaunt, tired, but with eyes that were burning with an intensity that scared people mostly my mother. My younger sister compared me to an anorexic model.</p><p>It was a surreal irony. I was spending 14 hours a day optimizing the memory hierarchy of a computer kernel to ensure zero starvation for CPU tasks while I was literally unintentionally starving my own biological cells. I was treating my body like a legacy system I could ignore, hoping the software would carry the load.</p><p>Then came COVID mid-build, and the brain fog hit. For a week, I couldn’t hold a complex C++ structure in my head. My hands were too shaky to type.</p><p>But I refused to stop. If I couldn’t write code, I would read it. I lay in bed, shivering, reading the history of the Linux kernel, and Neal Stephenson’s In the Beginning… was the Command Line. I used the sickness as an excuse to master the theory when I couldn’t practice the craft.It paid off</p><h4>The Southam 217 Sanctuary</h4><p>When I wasn’t sick, I wasn’t in my dorm. I was in <strong>Southam Hall, Lecture Room 217</strong>.</p><p>This became my monastery. At night, the room was empty. I would set up a workplace befitting of the work I did. It had two massive monitors, my laptop, and my tablet. I would cover the massive lecture hall blackboards with mind maps of kernel architectures, payoff matrices for <a href="https://fawazishola.ca/tenet/">Tenet</a>, and flowcharts for the <a href="https://fawazishola.ca/alexitha/">Alexitha</a> swarm. Sorry, cleaners who left the boards washed before I used them.</p><p>There is a distinct pleasure in being the only human in a space designed for hundreds. The silence of Southam Hall at 3:00 AM is heavy. It presses on you. It forces you to confront the quality of your own thoughts because there is nowhere to hide, and the use of a blackboard slowed the thoughts down enough to be deep.</p><h4>The “Mother’s Telemetry”</h4><p>While I was building a “Zero-Telemetry” OS to escape surveillance, I was living under the most intense surveillance system known to man: <em>my mother.</em></p><p>She knew about the fainting spells in high school. She saw the weight loss. She knew the cost of my obsession. Her worry wasn’t “helicopter parenting”; it was a data-driven response to a history of system failure.</p><p>She instituted an “Energy Drink Ban.” She required a daily “Check-In” protocol that felt like a heartbeat sensor.</p><p><strong>The Climax: 19 Missed Calls</strong></p><p>One time, I went deep. I was debugging a race condition in the Nash Scheduler. I put on my headphones, locked into the code, and let the world dissolve. Time ceased to exist.</p><p>When I finally surfaced, hours later, I looked at my phone.</p><p>19 Missed Calls.</p><p>Mom. Dad. Sister. Campus Security.</p><p>My mother hadn’t heard from me. She knew my history. She panicked. She had alerted Campus Security and the residence staff to hunt me down because I had “gone dark” for too long.</p><p>Seeing those notifications was a shock. I realized that while I was managing microsecond jitter in a kernel, I had failed to manage the heartbeat of the people who loved me. I was so obsessed with digital safety that I had become a physical liability.</p><p><strong>Patching the Family Interface</strong></p><p>There is an upside to spending your life architecting kernels: you start seeing easy technical solutions for messy human problems.</p><p>To prevent another “19 Missed Calls” panic, I spent an afternoon building <strong>Pulse</strong>. It’s a simple telemetry dashboard installed on my family’s phones. It gives them a live read on my status <strong>Red for Busy</strong>, <strong>Green for Free</strong>. It logs my last meal and my sleep state. Now, my mother doesn’t have to call 19 times to know I’m alive; she just checks the dashboard. She gets the data she needs, and I get the silence I need.</p><h4>The Champagne Birthday: The 19km Proof</h4><p>On June 19th, I turned 19. It was my “Champagne Birthday.”</p><p>I was at my physical weakest. I had lost the muscle. I was still recovering from the academic year. Most people would have thrown a party.</p><p>I decided to run 19 kilometers.</p><p>I had zero training. I hadn’t run a mile in six months. My body was not ready. But my mind needed to know something.</p><p>I needed to prove that “hard” was my baseline. I needed to prove that my will was stronger than my biology.</p><p>I laced up my shoes and I ran.</p><p>The first 5k was fine. The next 5k was painful. The last 9k was a conversation with every cell in my body begging me to stop. I ignored them. I treated the pain as “noise” and filtered it out.</p><p>When I finished, I couldn’t walk properly for a week. But I felt a euphoria that no party could provide.</p><p>I realized something interesting my physical abilities. Knowing I <em>could</em> do a half-marathon on a broken body meant I could do anything if I actually trained. It wasn’t about the run; it was about the override.</p><h3>X. COGNITIVE CONDITIONING: THE PAVLOVIAN PROTOCOL</h3><p>People ask me how I maintain this intensity without burning out. The answer is simple: I don’t rely on “motivation.” Motivation is a feeling, and feelings are non-deterministic. Instead, I rely on <strong>Conditioning</strong>.</p><p>I have treated my own brain like a neural network that needs to be trained on specific inputs to produce specific outputs. I have engineered a physical environment that forces me into “Flow State” instantly.</p><h4>The Three-Headset System (Hardware Separation of Concerns)</h4><p>Just as Axiom uses tiered memory (RAM vs. VRAM) for different tasks, I use tiered hardware for my brain. I don’t just switch tasks; I switch physical interfaces.</p><p><em>Tier 1: The Deep Work Lock down</em></p><p>When I need to write kernel code or derive a new proof for <a href="https://fawazishola.ca/flux/">Flux</a>, I reach for a specific setup: Wired headphones plugged into a 1st Generation iPod Shuffle.</p><p>Why the Shuffle? It is the “Axiom” of music players. No screen. No internet. No Bluetooth latency. It is a hardware commitment to a single task.</p><p>The soundtrack is always the same: <strong>Hans Zimmer’s Interstellar Theme</strong>. When the organ kicks in, the scale of the math in my head becomes infinite. The music triggers a Pavlovian response, my heart rate slows, my focus gets sharper, and the code flows, hours fly by. It is the soundtrack of the deep work.</p><p><em>Tier 2: The Busy Work Flow</em></p><p>For administrative tasks, emails, writing,or light testing, I switch to a pair of light J Lab overhead headphones. These feel different on my head. They are a tactile signal to my brain: “We are productive, but we are reachable.”</p><p>The soundtrack shifts to <strong>Drake on repeat</strong>. It provides a rhythmic, cultural backdrop that keeps me moving without demanding deep cognitive processing. This is for the “Lanes.”</p><p><em>Tier 3: The Life Boundary</em></p><p>I have a third, separate pair of headphones for the gym or walking to class. I never use them for work. This prevents “leakage.” My brain knows exactly what “Instruction Set” to load based on the weight of the plastic on my ears.</p><p>I have reduced the “context-switching cost” of my own psychology to near-zero.</p><h3>XI. THE AUTODIDACT’S IDENTITY: EFFORT OVER INNATE ABILITY</h3><p>I am often called a “Polymath.” I don’t refute the label, but I don’t think it captures me fully. To me, it feels static. It implies a state of <em>being</em> rather than a state of <em>doing</em>.</p><p>I prefer to think of myself as a <strong>Curious Autodidact</strong>. I am simply someone who refuses to accept that two fields are separate just because they are taught in different buildings.</p><h4>The Case Study: The F1-Quant Engine (Age 17)</h4><p>This instinct didn’t start at university. At 17, while other kids were memorizing driver stats, I was trying to build a non-linear prediction model for the stock market based on Formula 1 telemetry.</p><p>I looked at the logistics of the sport , the tire supply chains of Pirelli, the shipping routes of DHL, and the tech sponsorships of Oracle and Petronas. I realized that F1 wasn’t just a race; it was a high-speed stress test for the global industrial economy.</p><p>If I could track the micro-movements of a pit stop or a supply chain delay in the paddock, I believed I could predict macro-movements in the automotive and logistics sectors before Wall Street algorithms caught up. I was looking for the “invisible wires” that connected a race car in Monaco to a stock ticker in New York.</p><p>That project taught me that everything is connected. “Polymathy” isn’t about knowing many things; it’s about seeing the threaded connections that others miss.</p><h4>The “Exclusive Space” Challenge</h4><p>I fell in love with Formula 1, a sport where I rarely saw faces like mine in the paddock or the engineering bays. But instead of feeling excluded, I felt challenged.</p><p>I realized that the language of aerodynamics and the logic of game theory are universal. If I could speak the language of “<strong>First Principles</strong>” better than anyone else, my identity wouldn’t be a barrier; it would be a distinct advantage. I didn’t want to just be “invited” to the table; I wanted to build the table.</p><h4>Redefining Genius</h4><p>I reject the idea of “innate genius.” Being praised for being “smart” feels like an insult to the work. It wipes out the 14-hour days, the gastritis, and the 3:00 AM resets when the kernel panics for 20 times consecutively.</p><p>I define genius differently. <strong>Genius is the audacity to take action.</strong> It is the willingness to spend 100 hours solving a problem that others give up on in ten. I would rather be known as the most hardworking person in the room than the smartest.</p><h3>XII. THE MUSEUM OF FAILURE: THE /MORGUE DIRECTORY</h3><p>I maintain a directory on my hard drive simply titled /morgue. It is not a trophy case. It is a graveyard.</p><p>Inside, you will not find Axiom or Flux. You will find the corpses of the 31 unique boot failures that panic-looped the kernel. You will find “The Symlink Crisis,” where a blind copy of a library file pointed to a non-existent directory, causing the entire AI stack to hallucinate.</p><p>I do not delete these failures. I archive them.</p><p>To the outside world, I am the 19-year-old claiming he can rewrite the computing stack. They see the “Innate Genius” or the “audacity.” <strong>But they are falling for the Survivorship Bias of the demo.</strong> They see the one version of the system that booted. They miss the 400 versions that segfault-ed.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XNg0o94v0Kesj81dP3Cz7g.jpeg" /><figcaption><strong>Failure #14.</strong> The kernel panicked and “Attempted to kill init” effectively murdering the operating system from the inside.</figcaption></figure><h3>XIII. THE 15-YEAR TIME CAPSULE (MESSAGE TO 2041)</h3><p>I started writing this on a cold night in Ottawa in 2025.</p><p>If you are reading this in 2041, you likely know me as something else.I am a 35 year old with more life experience. You know “<a href="https://fawazishola.ca/axiom/"><strong>Axiom</strong></a>” , or “<strong>Atlas</strong>” . You see the outcome.</p><p>But I want you to see the origin.</p><p>I am writing this to freeze this moment in time. I want there to be a permanent record that this didn’t start in a boardroom. It didn’t start with a venture capital check. It started with a teenager who nuked his hard drive because he refused to live in a box.</p><p>This document is the <strong>Genesis Block</strong> of FAWAZ. It is the proof that the philosophy came before the capital.</p><p>To the reader of the future: Do not let the scale of what I built distract you from the simplicity of how it started. It started with a refusal to accept the status quo. It started with the 4KB Spirit.</p><p>If you are reading this and feeling trapped in your own “Lane,” let this be your permission to swerve. Nuke your own drive. Build your own map. The world is infinitely malleable to those who refuse to accept it as it is.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=afbfecf43301" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Power of One]]></title>
            <link>https://fawazishola.medium.com/the-power-of-one-f498dd6298a5?source=rss-82891d5fee70------2</link>
            <guid isPermaLink="false">https://medium.com/p/f498dd6298a5</guid>
            <category><![CDATA[systems-thinking]]></category>
            <category><![CDATA[stem]]></category>
            <category><![CDATA[mathematics]]></category>
            <category><![CDATA[education]]></category>
            <category><![CDATA[society]]></category>
            <dc:creator><![CDATA[Fawaz]]></dc:creator>
            <pubDate>Fri, 23 Jan 2026 23:18:08 GMT</pubDate>
            <atom:updated>2026-02-08T18:24:47.872Z</atom:updated>
            <content:encoded><![CDATA[<p>Imagine a classroom. The air is thick with the scent of dry-erase markers and the low hum of cooling fans. For most students, the “potential energy” of the room is static: a set of rules to be followed and a curriculum to be absorbed.</p><p>But watch what happens the instant a Black teacher walks through the door to lead a physics lecture.</p><p>The shift is immediate, yet nearly invisible to the untrained eye. It isn’t just a change in personnel or a simple nod to diversity; it is a fundamental recalibration of the room’s physics. This is the proposition: the radical, disproportionate, and asymmetrical power of “Just one more.”</p><p>In social architecture, we often view impact as additive; one person plus one person equals two units of progress. But the Math of Influence follows a different logic: <strong>n × 10ˣ</strong>. Impact is exponential, not linear. That teacher’s primary function isn’t merely the delivery of mechanics; it is the validation of existence.</p><p>When a student sees the “One,” the mental path to mastery transforms from a theoretical impossibility into a proven trajectory. This is the <strong>Cascade Effect</strong>. One student sees the One, realizes that the path is compatible with their own “hardware,” achieves it, and eventually mentors ten more.</p><blockquote>Representation is, at its core, the replication of competence across a network.</blockquote><p>Without that initial node to anchor the signal, the signal never propagates. It dies in the silence of isolation.</p><p>I grew up in Nigeria, an environment I now call <strong>Bare Metal</strong>.</p><p>In an all-Black country, excellence was not an exception; it was the default setting. You didn’t have to wonder if a Black man could be a neurosurgeon, a commercial pilot, or a lead software architect. You saw them every day at the grocery store.</p><p>In this “Native Environment,” there was no identity friction. I never had to ask permission to be smart. My mental model was robust and sharp because it was reinforced by the architecture of my society: <em>“I never doubted I could do it because everyone did it.”</em></p><p>In another life, I would have probably pursued the path of neurosurgery. The idea never strays far from the back of my mind, all because of my mother. At age 9, she gifted me two books by Dr. Ben Carson,<strong> a world-renowned Black Neurosurgeon</strong> who overcame humble beginnings to lead the first successful separation of <strong>Craniopagus</strong> twins.</p><p>Reading those books, I saw a great deal of myself in him. Just like him, I despised studying. We both had mothers who forced us to read more. That level of relatability stuck deep in my brain; it provided an unwavering sense of confidence that I could go into that field if I ever wanted to.</p><p>I never once doubted if I would make it to medical school or survive the long hours. Having him as a representation solved the equation. I knew all I had to do was put in the work. <strong>(Though, sorry Mom, it looks like I’m not going to be getting a ‘Dr.’ before my name after all, unless, of course, I eventually go for a doctorate).</strong></p><p>The learning process in my native environment was highly optimized. 100% of my “compute power” was dedicated to the subject matter. When I sat down to learn, my brain wasn’t wastefully running background processes wondering if I belonged in the room. I was running <strong>Bare Metal</strong> , executing directly on the hardware without the overhead of a bloated social operating system.</p><p>Then, at age 12, I migrated to the West.</p><p>The transition felt like moving from a high-speed fiber connection to a 56k modem. I was enrolled in a highly evaluated STEM program, a prestigious track designed to produce the next generation of engineers and scientists. My class consisted of over 120 students.</p><p>When I looked at the data later, the realization was staggering: <strong>exactly 0.8% of that STEM class was Black.</strong></p><p>That 0.8% was me. I was the anomaly.</p><p>The shock wasn’t just in the low number; it was in the discrepancy. The school’s general population was 13% Black. Yet, as the filters tightened and the curriculum shifted toward advanced science and mathematics, that 13% collapsed into 0.8%.</p><p>Somewhere between the hallways and the laboratory, the system had dropped the signal.</p><p>I find it revealing to examine what happens when the same “hardware” runs on different architectures. Consider two nodes in a distributed network. Both possess identical processing power (raw IQ) and identical memory specifications.</p><p><strong>Node A</strong> operates in a “Native Environment.” The architecture is optimized for its instruction set. It runs “Bare Metal.” 100% of its compute cycles are dedicated to the primary objectives.</p><p><strong>Node B operates in a “Hostile Environment.” Here, the system architecture treats the node as an exception. To function, Node B is forced to run a heavy virtualization layer, a “hypervisor” of social translation.</strong></p><p>This layer requires constant, invisible background processing. It manages exception handling (<em>“Do I belong here?”</em>), security protocols (<em>“Is this a microaggression?”),</em> and compatibility checks (<em>“Am I being understood?”</em>).</p><p>I remember a specific moment in gym class that perfectly illustrates this latency. I was complaining to another Black student about my utter disdain towards chemistry (sorry, Ms. V). He didn’t offer a commiserating shrug. Instead, he looked at me with pure, unadulterated confusion, as if I were describing a trip to Mars.</p><p>To him, the very idea of me taking chemistry was an error in the code. It wasn’t that he thought I wasn’t smart; it was that the system had convinced him that chemistry wasn’t a “valid” workload for people like us.</p><p>This is the cost of the Virtualization Layer. It is a massive allocation of resources to System Overhead. Node A is running at 100% capacity. Node B is being stripped for its thread count: 60% on computation, 40% on maintaining the virtualization layer.</p><p>This is not a difference in capability. It is a difference in <strong>Throughput Efficiency</strong>.</p><blockquote>We are running high-performance hardware on a bloated, unoptimized operating system, and then blaming the hardware for the latency.</blockquote><p>I find it useful to move this conversation away from the language of “Morality” and toward the language of <strong>Efficiency</strong>.</p><p>If we look at our civilization through the lens of <strong>System Architecture</strong>, the current state of STEM is not just a social injustice; it is a catastrophic technical failure.</p><p>It is no coincidence that a significant portion of Black people currently in Western STEM are immigrants. They often succeed because they began their “booting” process in environments where excellence was the default. They arrived with their “firmware” already established, allowing them to better withstand the latency of the Western system. But we shouldn’t require a “pre-boot” in another country for a mind to function here.</p><p>High-performance computing operates on a fundamental axiom: <strong><em>A system that leaves 90% of its resources idle is a failed system.</em></strong></p><p>When we look at the “Pipeline Problem,” we aren’t looking at a lack of raw talent. We are looking at a massive, systemic Resource Allocation Failure. By creating environments with high “Identity Friction,” we are effectively throttling the compute power of our species. We are trying to solve the world’s most complex problems,from climate modeling to AGI safety; while running on a fraction of our available “nodes.”</p><p>Utterly inefficient.</p><p>Think of the “Power of One” as the <strong>Initial Bootloader.</strong></p><p>In computing, the bootloader is that tiny, critical piece of code that initializes the hardware so the operating system can load. It acts as a catalyst. It doesn’t do the heavy lifting itself, but without it, the most powerful H100 GPU in the world is just a high-priced paperweight.</p><p>Representation is the ultimate social bootloader. It initializes the mind. It clears the “Identity Friction” and the “Outlier Bugs” so that the student’s internal operating system can finally load.</p><p>Once the bootloader has run, the mind is free to wander through unrestricted depths of its own choosing, be it Quantum Mechanics, fluid dynamics, or Distributed Systems; at its maximum clock speed. We are currently losing “flops” of human genius every single day because the bootloader never runs.</p><p>We must begin debugging society with the same level of intensity, lack of sentimentality, and commitment to <em>truth</em> that we use when debugging a technical system. The “Outlier Bug” is not a feature of a meritocracy; it is a flaw that crashes the system’s ability to innovate at scale.</p><p>If we want to solve the hardest problems of the next century, we cannot afford the latency of exclusion. We cannot afford the “Tax of the Only.” Our objective must be to optimize for maximum throughput, to ensure that every single individual node in our global network can run at its true, intended frequency.</p><p>The call to action is simple: <strong>Provide the bootloader. </strong>Not just to be “inclusive” or “diverse,” but to fix the architecture so the system can run at full power.</p><p>It’s way past time to patch the system.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f498dd6298a5" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Lanes are for traffic.]]></title>
            <link>https://fawazishola.medium.com/lanes-are-for-traffic-7193d676a4bd?source=rss-82891d5fee70------2</link>
            <guid isPermaLink="false">https://medium.com/p/7193d676a4bd</guid>
            <category><![CDATA[education]]></category>
            <category><![CDATA[polymath]]></category>
            <category><![CDATA[philosophy]]></category>
            <category><![CDATA[creativity]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[Fawaz]]></dc:creator>
            <pubDate>Fri, 16 Jan 2026 12:08:31 GMT</pubDate>
            <atom:updated>2026-01-16T12:08:31.298Z</atom:updated>
            <content:encoded><![CDATA[<h3>Brilliance Shouldn’t Be Restricted to a Single Lane</h3><p>We are taught from a young age to “pick a lane.” We are told that mastery requires a singular focus and that to wander is to be lost. But what if the lane we were assigned is too rigid for the vehicle we have? The truth is, brilliance restricted to a single lane is an aberration; rather, it requires the freedom of the entire map.</p><p>For a long time, I fell for the trap. I viewed myself as a “technical mind” who happened to have a capacity for creativity. I saw them as two separate rooms in my brain. It wasn’t until recently that I realized the fundamental flaw in that logic: <strong>To be human is to be creative.</strong> Whether I am writing code, solving equations, or writing poetry, it is all <em>creation</em>. It is all art; only the mediums and the tools change.</p><h3>The Myth of the Box</h3><p>Oscar Wilde famously said, <em>“To define is to limit.”</em> Yet, we live in a society obsessed with “labeling.” We want to put people in boxes so we know where to file them. We’ve even weaponized language to keep people in their place. We’ve all heard the warning: <em>“Jack of all trades, master of none.”</em></p><p>But we rarely hear the end of that quote:</p><blockquote><strong>“…but oftentimes better than a master of one.”</strong></blockquote><p>The omission is intentional. It’s a narrative designed to convince us that excellence is incapable of spanning multiple paths. History tells a different story.</p><h3>The Orchestration of the Intersections</h3><p>When I look at the “greats”—Leonardo da Vinci, Steve Jobs, Virgil Abloh—I don’t see specialists.</p><p>Steve Jobs is often misunderstood as a “tech” guy, yet he never coded the Mac or built the circuitry of the iPhone. His brilliance wasn’t in “doing” one specific technical task; it was in the <strong>knowing</strong>. He understood that technology need not be boring or tasteless, a mistake his competitors (Bill and IBM) made for decades and still make today by treating tech as a utilitarian chore.</p><p>Jobs led the most valuable company in the world by simply believing in the marriage of the arts and technology. He didn’t just build products; he led an orchestra. He knew that the magic happens at the intersection of the lanes. By refusing to be placed in the “technology box,” he was free to see the harmony between a calligraphy class and a computer operating system. We don’t need more specialists; we need more conductors.</p><h3>The Renaissance Individual in the AI Age</h3><p>This synthesis is no longer just a luxury of the “greats”; it is becoming a survival requirement. We are entering an era where artificial intelligence can out-specialize any specialist. If value is tied to a single, narrow lane, we are competing with a machine that never sleeps and processes data at the speed of light.</p><p>The machine is the ultimate specialist, but it is confined to the lane of its training data. It depends on us for its input.</p><p>The human’s greatest edge is the ability to drive across the whole map while the machines stay in their lanes. AI can calculate the structural integrity of a wing, but it cannot yet explain how the aesthetics of that wing’s curve might evoke a psychological sense of freedom that shifts a culture’s political leanings. True synthesis: the ability to bridge the “why” of philosophy with the “how” of engineering still remains a territory that belongs uniquely to humans. In the AI age, the most valuable skills are the ones that can help humans connect the dots that the machine doesn’t even know exist.</p><h3>The Universal Language of the Map</h3><p>As I study engineering, I find myself increasingly frustrated by the “lanes” of academia. To me, these aren’t isolated subjects; they are different dialects of the same language.</p><p>I want to be able to talk about how the mathematical models used in economics can be applied to population dynamics in biology. I want to explore the anthropological reasons our brains evolved the way they did to dictate the foundation of modern politics. I want to dissect the philosophy of the rules that govern us and realize how arbitrary they often are, just like the definition of money itself.</p><p>Money only has value because we say it does. It is a social fiction, a piece of engineering that only works if we all believe in the same design. A foray into its history and evolution over time exists at the convergence of a multitude of fields that’d seem separate but come together to form pillars of this social construct.</p><p>When you start to see that the logic of a rocket engine shares a heartbeat with the logic of an economic system or the evolution of a species, the “lanes” disappear. You realize that “social engineering” and “aerospace engineering” aren’t just sharing a word; they are sharing a fundamental quest to understand how things work under pressure and the underlying mechanisms to be triggered to effect certain outcomes.</p><h3>Tearing Down the Walls</h3><p>To define yourself is to place yourself in a box, and boxes are where things go to stay still.</p><p>I am choosing to stay in motion. I am choosing to be “UN-labeled.” By refusing to stay in a single lane, I am not losing my way; I am expanding my territory.</p><p>We need to stop asking children what they want to <em>be</em> and start asking them to dive deep into what areas interest them and in what ways they converge. Because when you understand the map, you realize you don’t need a lane. You have the whole world.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7193d676a4bd" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The $Trillion Economy No One Is Building For]]></title>
            <link>https://fawazishola.medium.com/the-trillion-economy-no-one-is-building-for-6c21ee4aa79a?source=rss-82891d5fee70------2</link>
            <guid isPermaLink="false">https://medium.com/p/6c21ee4aa79a</guid>
            <category><![CDATA[economics]]></category>
            <category><![CDATA[creativity]]></category>
            <category><![CDATA[future-of-work]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Fawaz]]></dc:creator>
            <pubDate>Fri, 16 Jan 2026 11:56:09 GMT</pubDate>
            <atom:updated>2026-01-16T11:56:09.621Z</atom:updated>
            <content:encoded><![CDATA[<h3>The $3 Trillion Economy No One Is Building For</h3><p>In grade 11 gym class, I watched a documentary about modern farming practices that stuck with me for years. One segment showed how today’s chickens reach market weight in 47 days, half the time it took chickens in the 1950s, while growing 2.5 times larger. But there was a gruesome problem: their legs couldn’t keep up. Nearly 30% of these birds suffered from broken bones and lameness. Their biological infrastructure was collapsing under the weight of growth that evolution hadn’t prepared them for.</p><p>Years later, staring at the projections for the creator economy: $104 billion in 2020, $250 billion today, and a projected $3 trillion by the 2030s. I made the visceral connection.</p><p>We are watching an economic sector grow 30x in just over a decade. And just like those chickens, the infrastructure supporting it wasn’t built for this kind of load. We are witnessing a skeletal failure of the modern economy. Creators are already breaking under the weight of managing multiple platforms, stitching together payment systems, building audiences on rented land, and cobbling together tools never designed to scale with them.</p><p>The economy is exploding. The infrastructure is snapping.</p><h3>The Anthropological Constant: Why This Isn’t a Bubble</h3><p>Skeptics call the creator economy a “bubble” or a “COVID hangover.” They are missing the historical data.</p><p><strong>Creativity is the pinnacle of human evolution.</strong> It is the one luxury that has never stopped being essential. Across every civilization, every economic system, and every crisis, humans have paid for distraction, paid for stories, and paid for spectacle.</p><ul><li><strong>The Romans</strong> had their Colosseum.</li><li><strong>Medieval courts</strong> had their jesters and bards.</li><li><strong>The Great Depression</strong> had its radio shows and the Golden Age of Hollywood.</li></ul><p>We have creators.</p><p>The platforms change. The economics shift. But the fundamental transaction of trading resources for entertainment and connection has remained constant for thousands of years. This is not a market trend; it is human nature.</p><p>As AI accelerates the automation of cognitive and manual labor, we are speeding toward a future where “survival work” (farming, manufacturing, data entry) is handled by machines. What remains for humans? <strong>Creation.</strong></p><p>We already create right now, either for work or on the side. When automation frees up 40 hours a week for billions of people, they won’t just sit in silence. They will move in droves to creation as entertainment. The demand for human stories, human humor, and human perspective will skyrocket because it will be the only thing machines cannot and will never perfectly replicate.</p><p>The “Creator Economy” isn’t a niche sector; it is the default economy of the AI-accelerated post-labor world.</p><h3>The Macro-Economic Disconnect</h3><p>To understand the scale of this failure, you have to look at the numbers. The creator economy is currently valued at roughly $250 billion. By 2027, Goldman Sachs projects it will reach $480 billion. By the mid-2030s, as AI displacement accelerates and traditional employment contracts decline, aggressive estimates place the total economic activity of independent creators, solopreneurs, and micro-media empires at nearly $3 trillion.</p><p>For context, a $3 trillion economy is just shy of the GDP of the United Kingdom. It is an economic entity that rivals the G7 nations.</p><p>Yet, this “nation” of creators is operating on financial and professional rails built for the 1950s. Our current banking systems, tax laws, professional networking sites, and payment processors were designed for the <strong>W-2 era. </strong>They were built for employees who receive a predictable biweekly paycheck from a single employer.</p><p>When a creator earning $500,000 a year from five different revenue streams applies for a mortgage, the system views them as “high risk.” Why? Because the algorithm cannot parse nonlinear income. The infrastructure views complexity as instability. This isn’t just a glitch; it’s a structural deficit. We have a 21st-century workforce trying to run on 20th-century rails, and the friction is costing us billions in lost productivity.</p><h3>The AI Forcing Function</h3><p>This isn’t just about YouTubers and influencers anymore. The rise of generative AI is acting as a massive forcing function, pushing millions of knowledge workers out of traditional employment and into the creator economy whether they want to be there or not.</p><p>As AI automates cognitive labor like copywriting, coding, data analysis, and basic design, the “safe” middle-class job is evaporating. The only economic refuge left is human-centric creation: distinct personality, unique taste, and community leadership.</p><p>This means we are about to see a migration of labor on the scale of the Industrial Revolution. Millions of people will transition from being “employees” to being “enterprises of one.” But unlike the farmers who moved to factories, these new digital workers are moving into a barren landscape. There is no union, no guild, no standardized benefits, and no dedicated banking system waiting for them. They are being pushed into the wilderness with nothing but a smartphone and a Stripe account. This is the reality in a supposedly more connected society</p><h3>The Tooling Fragmentation</h3><p>The current state of “creator tools” is a fragmented mess. A professional creator today has to manage a tech stack that rivals a small software company:</p><ul><li><strong>Identity:</strong> LinkedIn (built for corporate employees with rigid boxes), Instagram (built for ads), and Behance (static portfolios devoid of soul).</li><li><strong>Workflow:</strong> Notion, Trello, Slack, Discord, and Google Drive.</li><li><strong>Monetization:</strong> Patreon, Substack, brand deals, AdSense, and merch stores.</li><li><strong>Finance:</strong> QuickBooks, PayPal, separate invoicing tools.</li></ul><p>These tools are an insult to the creator class. Each of these tools is an island. None of them talk to each other. A creator spends 60–70% of their time acting as the IT administrator, CFO, and HR manager of their own life, leaving only 30% for actual creation. It doesn’t take a mathematician to model the amount of value that evaporates into thin air daily for this class of workers. These tools require them to wear multiple hats but put them in boxes to define themselves</p><p>This is the definition of inefficient infrastructure. In the physical world, we don’t ask truck drivers to pave their own roads before they can deliver goods. Yet, in the digital economy, we ask creators to build their own operational stack before they can earn a dollar, if they earn a dollar.</p><p>The fracture is visible everywhere. Creators lose data when platforms change algorithms. They lose income when payment processors freeze funds without explanation. They lose momentum because they are trapped in “admin hell.”</p><h3>The Opportunity Cost of Friction</h3><p>The cost of this friction isn’t just frustration; it’s lost GDP.</p><p>When a creator spends 4 hours a day managing tools instead of creating, that is 50% of their productive capacity wiped out by bad infrastructure. Across a workforce of 50 million creators (and growing), that amounts to hundreds of billions of dollars in unrealized value every single year.</p><p>We are leaving a trillion dollars on the table because we refuse to build the proper roads.</p><p>The next great technology company won’t just be another “tool” for creators. It won’t be another newsletter platform or another link-in-bio app. It will be the <strong>Infrastructure Layer</strong> itself. It will be the company that unifies identity, finance, and workflow into a single, cohesive operating system for the individual enterprise.</p><p>We don’t need faster chickens. We need a stronger skeleton.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6c21ee4aa79a" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>