Definition: The extent to which an AI’s goals are in line with its creators’ goals, ensuring safe and beneficial AI behaviour. What is Alignment in Coding? In broad AI safety, Alignment means “don’t destroy humanity.” In Vibe Coding, Alignment means “don’t destroy my repo.” It’s about ensuring the AI adopts Your Style, Your Conventions, and…
Definition: Output from generative AI systems favouring quantity over quality, such as low-quality, cheaply produced AI-generated content. Identifying “Slop” in Code AI Slop is code that looks like code, runs like code, but is actually garbage. The Vibe Coding Danger The speed of vibe coding makes Slop dangerous. You can generate 1,000 lines of Slop…
Definition: The capability to monitor, understand, and trace AI system behaviour, outputs, and decision-making processes. Why You Need It When you write traditional code, you use a debugger to step through lines. When you “vibe code” with AI, the logic is hidden inside the model’s weights. AI Observability is the set of tools and practices…
Definition: A dynamic process where an agent autonomously plans and executes actions to achieve goals, involving reasoning, invoking external tools, and self-correction. The “Plan-Act-Observe” Loop An Agentic Workflow is the specific pattern of behavior an AI follows to solve a problem. It’s not just “one shot” generation; it’s a cycle. Why This Matters for Vibe…
Definition: AI systems designed to autonomously pursue complex goals and workflows with limited direct human supervision, breaking tasks into steps and adapting based on outcomes. What Defines “Agentic”? Traditional software is “passive”—it waits for you to click a button. Agentic AI is “proactive.” It has a goal, a loop, and tools. The Agentic Workflow in…
Definition: Advanced feature in tools like Cursor and Windsurf where AI actively searches and understands entire codebases for context-aware assistance. What is Agent Mode? Standard AI chat windows are “myopic”—they only see what you paste into them or the current open file. Agent Mode (or “Composer” / “Agent” tab in tools like Cursor) breaks this…
Definition: A sophisticated gradient descent algorithm that rescales gradients for each parameter, effectively providing independent learning rates. Adaptive Gradient Descent Explained AdaGrad (Adaptive Gradient Algorithm) was a breakthrough because it realized that not all parameters need to learn at the same speed. The Metaphor for Coding Workflows While AdaGrad is an internal mathematical optimizer, its…
Definition: A training approach where algorithms selectively choose which data to learn from, valuable when labelled examples are scarce or expensive. The Concept In traditional ML, you throw all data at the model. In Active Learning, the model is smart enough to say, “I already know what a cat looks like. Show me more examples…
Definition: Mathematical functions enabling neural networks to learn nonlinear relationships between features and labels, including ReLU, sigmoid, and tanh. What Are They? In a neural network, an activation function is the mathematical “gatekeeper” at the end of a neuron. It decides whether the neuron should “fire” (activate) or not based on the input. Without them,…
Definition: The process of removing physical, spatial, or temporal details from objects or systems to focus on other aspects of interest. The Evolution of Coding Abstraction In the early days, we managed memory addresses (Assembly). Then we managed variables (C). Then objects (Java). Then functions (Python). Vibe Coding represents the next layer of abstraction: Natural…