Show description
THE KACHE MANIFESTO // VIBES BASED ENGINEERING
THE KACHE MANIFESTO // VIBES BASED ENGINEERING
__ __ ____ __ __ __ ___
| | | / | / ]| | | / _]
| | || o | / / | | | / [_
| _ || | / / | _ || _]
| | || _ |/ \_ | | || [_
|__|__||__|__|\ ||__|__|| |
\____| |_____|
THE KACHE MANIFESTO
SPEED // INTUITION // SILICON
STATUS: COMPILING VIBES...
01_PHILOSOPHY
02_THE_AI
03_THE_SILICON
04_BUILD_IT
05_GLOSSARY
01_The Philosophy: "Vibes Based Engineering"
"I don't do hyperparameter sweeps. I just pick them myself... based on vibes. I've read the research papers. I know what I'm doing."
- Kache
The Anti-Academic Approach
Modern engineering—especially at big tech companies (McDonald's)—is paralyzed by "process." They run grid searches (testing 1000 settings) to find the "perfect" number. They hold meetings. They wait.
The Kache Method is about High Agency and Internalized Knowledge.
Intuition over Brute Force: If you truly understand the math (you read the papers), you don't need a computer to guess the settings for you. You know the learning rate should be 3e-4. You feel it.
The "Fuck It" Loop: Don't wait for permission. Don't wait for the perfect plan. Start the training run. If it fails, you learn why instantly.
Speed is Quality: A fast iteration loop (trying 10 things in a day) beats a slow "perfect" loop (trying 1 thing in a week).
"My Shit is Stable"
When he says this, he refers to the stability of his algorithms. In Reinforcement Learning (RL), training often collapses (the robot spins in circles and learns nothing). He has tuned his code so aggressively that he can throw anything at it—randomization, quantization, noise—and it still learns. He has removed the fragility.
02_The AI Stack: Sim2Real & Vision
"Rendered pixels... 100,000 triangles per room... procedurally generated worlds. Is it going to generalize? Of course it's going to generalize."
The Core Problem: Generalization
You want to train a robot to walk or fly. You can't do it in the real world because it takes too long and the robot breaks when it falls. You must use a Simulation (like a video game).
The Gap: The real world is messy. Simulations are clean.…
THE KACHE MANIFESTO // VIBES BASED ENGINEERING
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>THE KACHE MANIFESTO // VIBES BASED ENGINEERING</title>
<style>
:root {
--bg-color: #050505;
--text-color: #e0e0e0;
--accent-color: #00ff41; /* Hacker Green */
--accent-dim: #008F11;
--alert-color: #ff3333;
--font-main: 'Courier New', Courier, monospace;
--scan-line-color: rgba(0, 255, 65, 0.04);
}
* {
box-sizing: border-box;
scrollbar-width: thin;
scrollbar-color: var(--accent-color) var(--bg-color);
}
body {
background-color: var(--bg-color);
color: var(--text-color);
font-family: var(--font-main);
margin: 0;
padding: 0;
line-height: 1.6;
overflow-x: hidden;
font-size: 16px;
}
/* CRT Effect Overlay */
body::before {
content: " ";
display: block;
position: absolute;
top: 0;
left: 0;
bottom: 0;
right: 0;
background: linear-gradient(
to bottom,
var(--scan-line-color) 50%,
rgba(0, 0, 0, 0) 50%
);
background-size: 100% 4px;
z-index: 999;
pointer-events: none;
opacity: 0.6;
}
.container {
max-width: 900px;
margin: 0 auto;
padding: 20px;
border-left: 2px solid var(--accent-dim);
border-right: 2px solid var(--accent-dim);
min-height: 100vh;
position: relative;
z-index: 1;
}
header {
text-align: center;
border-bottom: 2px solid var(--text-color);
padding-bottom: 20px;
margin-bottom: 40px;
}
h1 {
font-size: 2.5rem;
text-transform: uppercase;
letter-spacing: 4px;
margin: 0;
text-shadow: 2px 2px 0px var(--accent-dim);
}
h2 {
font-size: 1.5rem;
border-bottom: 1px dashed var(--accent-color);
padding-bottom: 10px;
margin-top: 40px;
color: var(--accent-color);
text-transform: uppercase;
}
h3 {
font-size: 1.2rem;
color: #fff;
margin-top: 25px;
text-decoration: underline;
}
p {
margin-bottom: 15px;
color: #cccccc;
}
.quote-block {
background: #111;
border-left: 4px solid var(--accent-color);
padding: 15px;
font-style: italic;
margin: 20px 0;
color: #fff;
}
.quote-block cite {
display: block;
margin-top: 10px;
font-size: 0.8rem;
color: var(--accent-color);
text-transform: uppercase;
}
/* Glitch Effect */
.glitch {
position: relative;
}
.blink {
animation: blinker 1s linear infinite;
}
@keyframes blinker {
50% { opacity: 0; }
}
/* Navigation Tabs */
.nav-tabs {
display: flex;
justify-content: space-around;
border-bottom: 1px solid var(--accent-dim);
margin-bottom: 30px;
flex-wrap: wrap;
}
.nav-tab {
background: none;
border: 1px solid transparent;
color: var(--text-color);
padding: 10px 20px;
cursor: pointer;
font-family: var(--font-main);
text-transform: uppercase;
font-weight: bold;
transition: all 0.3s;
}
.nav-tab:hover, .nav-tab.active {
background-color: var(--accent-color);
color: black;
box-shadow: 0 0 10px var(--accent-color);
}
/* Content Sections */
.section-content {
display: none;
animation: fadeIn 0.5s;
}
.section-content.active {
display: block;
}
@keyframes fadeIn {
from { opacity: 0; transform: translateY(10px); }
to { opacity: 1; transform: translateY(0); }
}
/* Technical Lists */
ul.tech-list {
list-style: none;
padding: 0;
}
ul.tech-list li {
position: relative;
padding-left: 25px;
margin-bottom: 10px;
}
ul.tech-list li::before {
content: ">";
position: absolute;
left: 0;
color: var(--accent-color);
font-weight: bold;
}
/* Glossary Grid */
.glossary-grid {
display: grid;
grid-template-columns: repeat(auto-fill, minmax(280px, 1fr));
gap: 20px;
}
.term-card {
border: 1px solid #333;
padding: 15px;
background: #0a0a0a;
transition: border-color 0.3s;
}
.term-card:hover {
border-color: var(--accent-color);
}
.term-title {
color: var(--accent-color);
font-weight: bold;
margin-bottom: 5px;
display: block;
border-bottom: 1px solid #333;
padding-bottom: 5px;
}
/* ASCII Art */
.ascii-art {
font-size: 10px;
white-space: pre;
text-align: center;
color: #555;
margin: 20px 0;
overflow-x: hidden;
}
/* Part List Table */
table.parts-table {
width: 100%;
border-collapse: collapse;
margin-top: 20px;
font-size: 0.9rem;
}
table.parts-table th, table.parts-table td {
border: 1px solid #444;
padding: 10px;
text-align: left;
}
table.parts-table th {
background-color: #1a1a1a;
color: var(--accent-color);
}
/* Footer */
footer {
margin-top: 50px;
border-top: 1px solid #333;
padding-top: 20px;
text-align: center;
font-size: 0.8rem;
color: #555;
}
/* Code block */
.code-block {
background: #111;
padding: 10px;
border-left: 3px solid var(--alert-color);
font-family: 'Courier New', monospace;
color: #aaa;
margin: 10px 0;
overflow-x: auto;
}
.highlight {
color: var(--accent-color);
font-weight: bold;
}
/* Mobile Adjustments */
@media (max-width: 600px) {
h1 { font-size: 1.8rem; }
.nav-tabs { flex-direction: column; }
.nav-tab { width: 100%; text-align: center; border-bottom: 1px solid #333; }
}
</style>
</head>
<body>
<div class="container">
<header>
<div class="ascii-art">
__ __ ____ __ __ __ ___
| | | / | / ]| | | / _]
| | || o | / / | | | / [_
| _ || | / / | _ || _]
| | || _ |/ \_ | | || [_
|__|__||__|__|\ ||__|__|| |
\____| |_____|
</div>
<h1>THE KACHE MANIFESTO</h1>
<p>SPEED // INTUITION // SILICON</p>
<p class="blink">STATUS: COMPILING VIBES...</p>
</header>
<div class="nav-tabs">
<button class="nav-tab active" onclick="openSection('philosophy')">01_PHILOSOPHY</button>
<button class="nav-tab" onclick="openSection('ai-stack')">02_THE_AI</button>
<button class="nav-tab" onclick="openSection('hardware')">03_THE_SILICON</button>
<button class="nav-tab" onclick="openSection('build-guide')">04_BUILD_IT</button>
<button class="nav-tab" onclick="openSection('glossary')">05_GLOSSARY</button>
</div>
<!-- SECTION 1: PHILOSOPHY -->
<div id="philosophy" class="section-content active">
<h2>01_The Philosophy: "Vibes Based Engineering"</h2>
<div class="quote-block">
"I don't do hyperparameter sweeps. I just pick them myself... based on vibes. I've read the research papers. I know what I'm doing."
<cite>- Kache</cite>
</div>
<h3>The Anti-Academic Approach</h3>
<p>Modern engineering—especially at big tech companies (McDonald's)—is paralyzed by "process." They run grid searches (testing 1000 settings) to find the "perfect" number. They hold meetings. They wait.</p>
<p><strong>The Kache Method</strong> is about <strong>High Agency</strong> and <strong>Internalized Knowledge</strong>.</p>
<ul class="tech-list">
<li><strong>Intuition over Brute Force:</strong> If you truly understand the math (you read the papers), you don't need a computer to guess the settings for you. You <em>know</em> the learning rate should be 3e-4. You feel it.</li>
<li><strong>The "Fuck It" Loop:</strong> Don't wait for permission. Don't wait for the perfect plan. Start the training run. If it fails, you learn why instantly.</li>
<li><strong>Speed is Quality:</strong> A fast iteration loop (trying 10 things in a day) beats a slow "perfect" loop (trying 1 thing in a week).</li>
</ul>
<h3>"My Shit is Stable"</h3>
<p>When he says this, he refers to the stability of his algorithms. In Reinforcement Learning (RL), training often collapses (the robot spins in circles and learns nothing). He has tuned his code so aggressively that he can throw <em>anything</em> at it—randomization, quantization, noise—and it still learns. He has removed the fragility.</p>
</div>
<!-- SECTION 2: THE AI STACK -->
<div id="ai-stack" class="section-content">
<h2>02_The AI Stack: Sim2Real & Vision</h2>
<div class="quote-block">
"Rendered pixels... 100,000 triangles per room... procedurally generated worlds. Is it going to generalize? Of course it's going to generalize."
</div>
<h3>The Core Problem: Generalization</h3>
<p>You want to train a robot to walk or fly. You can't do it in the real world because it takes too long and the robot breaks when it falls. You must use a <strong>Simulation</strong> (like a video game).</p>
<p><strong>The Gap:</strong> The real world is messy. Simulations are clean. When you move the brain from the game to reality, it usually fails.</p>
<h3>The Solution: Domain Randomization</h3>
<p>Kache's approach is brute-force chaos.</p>
<ul class="tech-list">
<li><strong>Procedural Generation:</strong> He doesn't train the AI in one room. He writes code to generate <em>infinite</em> rooms with random walls, floors, and obstacles.</li>
<li><strong>Visual Randomization:</strong> He changes the lighting, textures, and colors every few seconds.</li>
<li><strong>The Result:</strong> The AI stops looking at "color of the floor" to figure out where it is (because the color keeps changing). It is forced to learn the <em>geometry</em> of the room. This makes it robust enough for the real world.</li>
</ul>
<h3>The Architecture: CNN + RL</h3>
<p>He mentions a "Convolutional Neural Network" (CNN). Here is the pipeline he is likely building:</p>
<div class="code-block">
INPUT [Camera Image] ---> <span class="highlight">CNN (Visual Cortex)</span> ---> [Features] <br>
INPUT [Sensor Data] ---> <span class="highlight">MLP (Body State)</span> ---> [State Vector]<br>
<br>
[Features + State] ---> <span class="highlight">PPO Policy (Brain)</span> ---> OUTPUT [Motor Commands]
</div>
<p><strong>Quantization:</strong> He mentions "doing a little quantization." This means converting the AI's math from 32-bit (heavy, slow) to 8-bit (light, fast). This makes the AI "dumb" mathematically, but allows it to run 4x faster on a small robot chip.</p>
</div>
<!-- SECTION 3: THE HARDWARE -->
<div id="hardware" class="section-content">
<h2>03_The Silicon: Hardware Speedrunning</h2>
<div class="quote-block">
"Design a stereo camera PCB... KiCad... 24 hour turnaround... It's a point and click adventure game."
</div>
<h3>The Workflow: KiCad -> JLC</h3>
<p>Designing electronics used to be a dark art. Kache treats it like a video game.</p>
<ol class="tech-list">
<li><strong>Schematic (The Logic):</strong> Draw how the chips connect (CPU pin 1 -> Sensor pin 4).</li>
<li><strong>Layout (The Art):</strong> Arranging the physical chips on the board and drawing the copper wires (traces). This is the "Point and Click Adventure."</li>
<li><strong>Fabrication (The Speed):</strong> He sends the files to a rapid prototyping house (likely JLCPCB in China). They build the board overnight and ship it via air.</li>
</ol>
<h3>The Sensors</h3>
<p><strong>Stereo Camera:</strong> Two cameras side-by-side. The computer measures the difference between the two images (parallax) to calculate how far away things are. It works just like human eyes.</p>
<p><strong>Time of Flight (ToF):</strong> A sensor that shoots a grid of invisible infrared lasers. It measures the time it takes for the light to bounce back. This gives you a perfect "Depth Map" (a 3D image of the world) even in total darkness.</p>
</div>
<!-- SECTION 4: BUILD IT YOURSELF -->
<div id="build-guide" class="section-content">
<h2>04_Build It Yourself: The BOM</h2>
<p>You want to build what he's talking about? A robot that sees in 3D and learns via AI? Here is your shopping list.</p>
<h3>The Brain (Compute)</h3>
<table class="parts-table">
<tr>
<th>Component</th>
<th>Recommendation</th>
<th>Why?</th>
</tr>
<tr>
<td><strong>AI Computer</strong></td>
<td>NVIDIA Jetson Orin Nano (or Raspberry Pi 5)</td>
<td>You need GPU cores to run the Neural Networks. The Jetson is the gold standard for robotics.</td>
</tr>
<tr>
<td><strong>Microcontroller</strong></td>
<td>ESP32 or STM32</td>
<td>For low-level motor control. The Jetson thinks; the ESP32 moves the muscles.</td>
</tr>
</table>
<h3>The Eyes (Sensors)</h3>
<table class="parts-table">
<tr>
<th>Component</th>
<th>Recommendation</th>
<th>Details</th>
</tr>
<tr>
<td><strong>Stereo Camera</strong></td>
<td>IMX219 Dual Camera Module</td>
<td>Two Sony sensors on one board. Connects directly to Jetson/Pi.</td>
</tr>
<tr>
<td><strong>ToF Sensor</strong></td>
<td>VL53L5CX</td>
<td>An 8x8 zone laser ranging sensor. Gives you a matrix of distances. Cheap and fast.</td>
</tr>
</table>
<h3>The Software Stack</h3>
<p>Don't write everything from scratch. Stand on the shoulders of giants (then optimize based on vibes).</p>
<ul class="tech-list">
<li><strong>Training:</strong> <span class="highlight">Isaac Gym (NVIDIA)</span> or <span class="highlight">Mujoco</span>. These are the physics simulators where you generate the "worlds."</li>
<li><strong>PCB Design:</strong> <span class="highlight">KiCad 8.0</span>. Free, open source, professional grade.</li>
<li><strong>AI Framework:</strong> <span class="highlight">PyTorch</span>. Don't use TensorFlow. Kache implies PyTorch when he talks about research papers.</li>
</ul>
</div>
<!-- SECTION 5: GLOSSARY -->
<div id="glossary" class="section-content">
<h2>05_The Glossary</h2>
<div class="glossary-grid">
<div class="term-card">
<span class="term-title">HYPERPARAMETERS</span>
The "settings" for the AI brain before it starts learning. (e.g., Learning Rate, Batch Size). Choosing these wrong means the AI learns nothing.
</div>
<div class="term-card">
<span class="term-title">INFERENCE</span>
When the AI is actually *working* (predicting), not learning. "Running inference on the edge" means the robot is thinking for itself, not connected to a server.
</div>
<div class="term-card">
<span class="term-title">QUANTIZATION</span>
Compressing the AI. Turning 32-bit floating point numbers (0.123456789) into 8-bit integers (12). It reduces precision slightly but increases speed massively.
</div>
<div class="term-card">
<span class="term-title">SIM2REAL</span>
Simulation to Reality. The hardest problem in robotics. Making code that works in a video game work in the physical world.
</div>
<div class="term-card">
<span class="term-title">PCB (Printed Circuit Board)</span>
The green (or black) board that holds chips. "Spinning a board" means designing and ordering a new one.
</div>
<div class="term-card">
<span class="term-title">TRACES</span>
The copper wires embedded inside a circuit board. Kache calls dragging these around a "point and click adventure."
</div>
<div class="term-card">
<span class="term-title">EPOCH / STEP</span>
One cycle of learning. "3 billion steps" means the AI has tried to solve the problem 3 billion times in the simulation.
</div>
<div class="term-card">
<span class="term-title">CONVOLUTIONAL (CNN)</span>
A type of AI architecture designed specifically for processing grids of data, like images (pixels).
</div>
</div>
</div>
<footer>
NO WARRANTIES IMPLIED. IF YOU BRICK YOUR JETSON, THAT'S ON YOU. <br>
GENERATED FOR THE ASPIRING MANIC ENGINEER.
</footer>
</div>
<script>
function openSection(sectionId) {
// Hide all sections
const sections = document.querySelectorAll('.section-content');
sections.forEach(section => {
section.classList.remove('active');
});
// Deactivate all tabs
const tabs = document.querySelectorAll('.nav-tab');
tabs.forEach(tab => {
tab.classList.remove('active');
});
// Show specific section
document.getElementById(sectionId).classList.add('active');
// Activate specific tab
// Find the button that calls this function with this sectionId
const activeTab = Array.from(tabs).find(t => t.getAttribute('onclick').includes(sectionId));
if(activeTab) activeTab.classList.add('active');
}
</script>
</body>
</html>