Escaping the Chatbot Abyss: The Birth of a Website Factory

The server crashed while I was trying to explain CSS to an LLM. It was 3 AM, and my brain was a cauldron of caffeine and existential dread.

I have a graveyard in my head. A sprawling, overgrown cemetery of brilliant project ideas that died not from lack of vision, but from an overdose of friction. Each tombstone reads: “Killed by WordPress,” “Suffocated by CSS,” or “Lost to the Abyss of ‘Later.’”

I tried. God knows I tried. The traditional web development stack felt like fighting a dragon with a toothpick. WordPress was a bloated, database-driven monstrosity that demanded constant attention. Every update was a roll of the dice; every plugin a potential security nightmare. And the design… oh, the design. Fighting with CSS to perfectly align a pixel, spending hours tweaking responsive layouts, was a soul-crushing exercise in futility for a mind that craves impact, not pixel-perfection. My passion was content, ideas, systems—not the tedious alchemy of front-end aesthetics. The ideas kept coming, but the pipeline for getting them out into the world was a leaky, rusty sieve. They retreated into the obscure corners of my mental landscape, lost to the fog of “I’ll do it later.” This wasn’t just inefficiency; it was a personal tragedy. My brain, an idea-generating machine, was choking on its own output.

The Siren Song of “Vibe Coding” and the Prototype Trap

Then, a glimmer of neon on the horizon: the AI revolution. “Vibe Coding,” they called it. The promise was seductive: simply describe your desire to a chatbot, and poof—a website. My ENTP brain, ever the pattern-seeker, immediately saw past the individual website. “This,” I thought, “is not about building a website. This is about building a Factory.” A sprawling, automated machine capable of churning out custom websites for every stray thought, every fleeting obsession. Scalable. Efficient. Autonomous. The ultimate cure for the “too many ideas, too little time” syndrome. It was going to turn my ADHD into an unfair advantage, a digital forge where raw ideas would become steel.

The Reality Check: It was a disaster.

Oh, the honeymoon phase was glorious. The first five minutes with an AI chatbot were intoxicating. “Build me a minimal blog about quantum physics,” I’d type. Poof! A prototype, shimmering in its nascent glory. “Add a neon theme!” Done! It was fast, it was responsive, it was almost… magical.

But magic, I quickly learned, is an illusion. As soon as I asked for complexity, for iteration beyond the initial spark, the system buckled. The AI hallucinated. It created what I affectionately (now) refer to as “monsters of error”—spaghetti code interwoven with invisible dependencies, non-existent functions, and an utter disregard for logical file structure. Fixing one AI-generated bug would invariably spawn three more. I was trapped in a Sisyphean loop, debugging errors that the AI itself had manufactured, only for it to create new errors in its attempts to “fix” the previous ones.

The fundamental problem wasn’t the AI’s intelligence; it was the context. A chat window is an ephemeral void. My meticulously crafted prompts, the iterative feedback, the code snippets—they all vanished into the abyss of the session history. The AI was operating in a perpetual present, disconnected from its past actions or the broader architectural vision. I was building castles on quicksand, and every new prompt was merely another gust of wind eroding the foundations. The promise of Rapid idea capture, structured content, minimal design, performance focus, automated error resolution, prototype to production scalability felt like a cruel joke. “Vibe coding” was a lie.

The Great Awakening: The CLI Agent as the Digital Cartographer

The breakthrough, as most paradigm shifts are, arrived unceremoniously at 3:00 AM (because when else does true clarity strike a night owl?). It wasn’t a grand revelation, but a pragmatic observation from a YouTube video demonstrating AI Agents operating directly within the Command Line Interface (CLI).

It clicked. The distinction was stark, brutal, and elegant:

• The Chat is for brainstorming. It’s the chaotic, free-associative thought process.
• The CLI Agent is for building. It’s the meticulous, persistent execution layer.

When you move the AI from the ephemeral chat window into the structured, persistent environment of the CLI, everything changes. The agent isn’t guessing; it’s reading. It gains access to the file system, to the Git repository. It can ls your directories, cat your existing code, and even grep for patterns. This constant, real-time awareness of the codebase transforms it from a disoriented, hallucinating chatbot into a veritable Digital Cartographer.

How the Cartographer Works:

Instead of a black box, the CLI agent becomes an extension of your own thought process, but with infinite patience and memory. When I ask it to, say, “Add a new content section for ‘The Synapse’,” it doesn’t just guess. It performs a sequence of verifiable actions:

1. Contextual Scan: It reads hugo.yaml for site structure, layouts/ for existing partials, and content/ to understand current sections.
2. File Generation: It executes hugo new content synapse/_index.md, committing that action to the file system.
3. Code Injection: It might then read layouts/_default/baseof.html or layouts/partials/header-mothership.html to suggest or implement the necessary Hugo templating changes.
4. Verification: It can even run hugo server or npx playwright test to ensure its changes haven’t introduced regressions.

This constant feedback loop, grounded in the physical reality of files and version control, dramatically reduces hallucinations. The “code mess” transforms into “version control,” easily reviewable and revertible. The agent operates not in an abstract void, but within the meticulously mapped coordinates of my project repository.

The Factory Stack: Laziness as a Virtue (and a Superpower)

Armed with this newfound understanding, I rebuilt the factory from the ground up. The design principles were simple: speed, simplicity, and absolute minimal cognitive load.

• Hugo: The static site generator is the engine. Its blazing speed means no databases to manage, no server-side processing to bottleneck. It renders pages faster than an idea can fully form in my head.
• Tailwind CSS: My old nemesis, CSS, is now an obedient servant. Instead of hand-coding styles, I describe the aesthetic to the AI agent, and it generates the utility classes. Design by prompt. Minimal human interaction. Max visual impact.
• Netlify: The deployment pipeline is the conveyor belt. Every git commit is a potential product launch. Changes are live instantly, reflecting the rapid-fire iterations of my mind. It’s the ultimate dopamine hit: immediate feedback, zero friction.

This isn’t just about coding; it’s about systematizing serendipity. It’s about building a digital extension of my brain, a rapid prototyping and publishing platform that finally keeps pace with my thoughts.

The Endgame: Launching All the Crazy Ideas (Before the Inevitable Boredom)

Let’s be brutally honest: I am more interested in building the machine—this Website Factory—than I am in any single website it produces. That, my friends, is the quintessential ENTP curse. The system, the elegant architecture, the sheer ingenuity of the automation… that’s the real thrill. The websites are merely the exhaust fumes of a beautifully humming engine.

My goal is audacious, as all good ENTP goals are. I want to scale this. I want to refine it into a fully autonomous system, capable of launching complex digital presences with minimal human intervention. I envision taking this to the Indonesian market first, offering a transformative solution for local businesses, then expanding globally.

And I need to do it fast.

Because the clock is ticking. The inevitable ENTP boredom will set in. And when it does, this factory needs to be self-sustaining, a monument to my fleeting genius, while I pivot to the next grand obsession. Apps. Customer service systems. AI-driven transport networks. The possibilities are truly endless.

But for now, at least, I have finally built the machine to capture them all. This is The Foundry, where raw ideas become digital steel, turning ADHD into an unfair advantage, and bridging the gap between “what if?” and “here it is.”

➜

The Foundry

Turning ADHD into an unfair advantage. The bridge between 'what if?' and 'here it is.'

Enter Portal