Wong Edan's
Cover
Artificial Intelligence

AI, SI, and MCP: The Unholy Trinity of Modern Tech

February 15, 2026 • By Azzar Budiyanto

The Madness of the Modern Machine: Why Your AI is Bored and Your Systems are Broken

Welcome to the digital asylum, my fellow keyboard warriors and late-night debuggers. If you are reading this, you are likely part of the “Wong Edan” tribe—those of us who have stared into the abyss of a terminal for so long that the blinking cursor has started whispering secrets about the future of the universe. We are living in a time where “Artificial Intelligence” is no longer a buzzword; it is the air we breathe, the coffee we drink, and the reason we haven’t slept since 2023. But today, we aren’t just talking about chatbots that can write mediocre poetry. We are diving into the visceral, systemic reality of Systemic Intelligence (SI), the Model Context Protocol (MCP), and the radical evolution of Skills required to survive this transition.

I’ve spent the last eight months in a caffeinated haze, building systems that make Claude scream internally. I’ve seen what happens when you treat an LLM not as a calculator, but as a product manager. It is messy, it is beautiful, and it is fundamentally systemic. If you think AI is just about prompts, you are still playing with sticks while the rest of us are building nuclear reactors in our basements. Let’s get weird.

The Evolution from AI to SI: Systemic Intelligence is the New God

The term “Artificial Intelligence” is actually quite limiting. It suggests a singular entity—a brain in a jar. But the real power of the current era lies in Systemic Intelligence (SI). SI is the shift from “How do I make this AI write code?” to “How do I build a system where AI, data, and human feedback loops interact to solve complex problems autonomously?”

What Actually is Systemic Intelligence?

Systemic Intelligence is the application of AI within a holistic framework. Think about the systematic reviews we see in high-stakes fields like colorectal cancer surgery or diabetes management. In these contexts, AI isn’t just a tool; it’s an enhancement of a systematic process. For instance, in surgical complications analysis, the AI doesn’t just “predict” outcomes; it analyzes the systemic interactions between postoperative data, patient history, and surgical techniques to provide a comprehensive risk profile. That is SI in action.

In the world of software, SI means moving away from “one-shot” prompts. We are building Slash Command Requirements Systems. Imagine a system where you don’t just ask Claude to “write a function.” Instead, the AI treats you as the product. It probes you, it analyzes the existing codebase, it checks for systemic vulnerabilities, and it refuses to move until the architectural integrity is sound. This is the difference between a parrot and a partner.

The Connectivity Crisis

Why do we need SI? Because AI systems are notoriously insecure and siloed. As noted in the 2026 International AI Safety Reports, risks aren’t just “malicious use”—they are systemic risks. When an AI fails, it doesn’t just give a wrong answer; it can cause a cascade of malfunctions across an integrated network. Systemic Intelligence is the attempt to build guardrails and “connective tissue” that make these systems robust rather than fragile.

MCP: The Model Context Protocol (The Glue We’ve Been Begging For)

If SI is the philosophy, then MCP (Model Context Protocol) is the plumbing. For the uninitiated, MCP is the standard that allows AI models to securely and seamlessly connect with data sources and tools. It is the end of the “copy-paste” era of AI development.

Breaking the Silos

Until recently, your AI was trapped. It didn’t know about your local files, it couldn’t see your database schema unless you fed it manually, and it certainly couldn’t interact with your enterprise tools without a messy, custom-built API nightmare. MCP changes the game by providing a standardized way for AI “clients” (like Claude or GPT) to talk to “servers” (your data, your tools, your environment).

“I built a system where Claude treats me as the product… it’s not about the AI’s intelligence, it’s about the context protocol that lets that intelligence touch the real world.”

Why MCP is a Developer’s Fever Dream

  • Universal Connectivity: With MCP, you can build a tool once and use it across any model that supports the protocol. No more rewriting tool-calling logic for every new LLM release.
  • Local Control: You can run MCP servers locally, giving the AI access to your specific project context without uploading your entire proprietary codebase to a third-party server.
  • Agentic Freedom: This is where the “Wong Edan” energy really kicks in. When an AI has a protocol to access your terminal, your browser, and your database, it ceases to be a chatbot and becomes an agent.

Consider the recent systematic analysis of vulnerabilities in agentic coding assistants. Many of these vulnerabilities exist because the “skills” and “tools” protocols are fragmented. MCP aims to standardize this ecosystem, making it harder for prompt injection attacks to hijack the systemic tools the AI has access to. It is the difference between giving a toddler a chainsaw and giving a master carpenter a specialized lathe.

The New Skillset: Moving Beyond the ‘Arcane Skill’ of Prompting

There is a dangerous promise at the heart of the AI boom: the idea that programming is no longer an “arcane skill” and that anyone can do it with a chatbot. This is a lie—or at best, a half-truth. While the barrier to entry is lower, the ceiling for High-Level Systemic Design has moved to the stratosphere.

Transversal Competencies and 21st-Century Skills

As AI handles the “syntax,” the human must handle the “system.” A systematic review of generative AI in education highlights that 21st-century skills are far broader than just “digital skills.” We are talking about:

  • Architectural Literacy: Understanding how the AI, the MCP, and the legacy database interact. You don’t need to know how to write a regex by heart, but you damn well better know why a regex is the wrong solution for a systemic data integrity problem.
  • Critical Inquiry: The ability to “interrogate” the AI. When the AI gives you a solution, can you spot the systemic hallucination? Can you see where the AI is being “lazy” and pushing a design defect that will haunt you in six months?
  • Adversarial Thinking: Given the rise in prompt injection and systemic risks, every developer needs to be a security researcher. You need to ask, “How could this tool be used to bypass my own safety protocols?”

The Andrew Ng Perspective: Stepping Up

Even Andrew Ng has pointed out that people with the “right AI skills” are given unprecedented opportunities to step up. But what are those skills? They aren’t just about knowing how to use a library. They are about Orchestration. In a world where AI can generate 10,000 lines of code in a minute, the human skill shifts to filtering, validating, and integrating that code into a systematic whole.

Security in the Age of SI: Why Everything is Broken

Let’s talk about the elephant in the server room: AI systems may never be fully secure. The very nature of an LLM—probabilistic, flexible, and context-dependent—makes it a nightmare for traditional security paradigms. This is where the “Wong Edan” personality becomes a survival trait. You have to be a little crazy to trust these things.

Prompt Injection and Agentic Vulnerabilities

When we give AI “Skills” (via MCP or other protocols), we open doors. A systematic analysis of vulnerabilities in agentic coding assistants has shown that prompt injection isn’t just about making a chatbot say something naughty. It’s about Remote Code Execution (RCE). If an AI agent has the “skill” to read an email and the “skill” to execute a terminal command, a malicious email can effectively take over your machine through the AI.

This is why Systemic Intelligence requires Systemic Security. We cannot just secure the model; we must secure the protocol (MCP) and the environment. We need to build “Air-Gapped Logic” where the AI can propose actions, but the systemic layer requires a “human-in-the-loop” or a secondary “security AI” to validate the transaction.

Case Study: The Surgical Precision of Systematic AI

Let’s look back at the research on colorectal cancer surgery. Why use AI there? Because the “system” of a hospital is incredibly complex. Complications aren’t just caused by a surgeon’s hand; they are caused by the interaction of pre-op prep, anesthesia, surgical tools, and post-op care.

An AI-enhanced systematic review allows doctors to see patterns that no human could. It identifies systemic weaknesses in the workflow. This is the ultimate goal of AI, SI, and MCP: to find the patterns in the chaos. Whether you are managing diabetes or a Kubernetes cluster, the goal is to move from reactive fixing to systemic optimization.

How to Build Your Own ‘Wong Edan’ AI Stack

So, how do you actually implement this without losing your mind? Here is the blueprint for a Systemic Intelligence stack:

Step 1: Implement the MCP

Stop writing custom API wrappers. Start using the Model Context Protocol. Set up an MCP server that exposes your local environment (filesystem, git, DB) to your agent. This creates a “shared brain” between you and the AI.

Step 2: Define ‘Skills’ as Atomic Tools

Don’t give the AI a general “fix my code” command. Define specific, atomic skills. fetch_documentation, run_unit_tests, check_security_vulnerabilities. By modularizing skills, you create a systemic boundary that makes the AI’s actions predictable and auditable.

Step 3: The ‘Slash Command’ Mentality

Adopt the /slash command requirements system. When you interact with your AI, treat it as a structured transaction.
/analyze --depth systemic --include security_vectors.
Force the AI to treat you as the product owner. It should ask you for clarifications, outline its plan, and wait for approval before touching the “system.”

Step 4: Embrace the Systematic Review

Use AI to review its own work and the work of other AIs. Build a loop where Model A writes the code, and Model B performs a “Systematic Analysis” of the vulnerabilities in that code. This multi-agent systemic approach is the only way to catch the hallucinations that fall through the cracks of a single LLM.

Conclusion: The Future Belongs to the Systemic Architect

The “arcane skill” of the future isn’t coding—it is Systemic Orchestration. We are moving into a world where AI is the muscle, MCP is the nervous system, and SI is the conscious mind. Your job, as a developer, a doctor, a manager, or a “Wong Edan” tech enthusiast, is to be the Architect.

We must acknowledge the risks—the malfunctions, the malicious use, and the systemic failures described in the safety reports of 2026. But we must also embrace the promise. The promise that we can build systems that are more than the sum of their parts. Systems that can help us cure diseases, build incredible software, and perhaps, finally, get a good night’s sleep while the AI handles the “internal screaming” for us.

Stay mad, stay brilliant, and keep building the system. The cursor is still blinking, and it’s waiting for your next command. Make it a systemic one.