Learning Hub
Learn AI: The Senior Practitioner's Reference
Twelve modules across four tracks. The concepts, architectures, and workflows that public-sector leaders need before approving AI initiatives.
Four tracks
Track 1
Foundations
What these systems are and how they learn.
Track 2
Context and Grounding
How to make models answer from trusted sources.
Track 3
Agents and Tooling
How models take action and interoperate.
Track 4
Applied Practice
How to deliver with AI in real engineering work.
All modules
Search by title, concept, or tag.
Generative AI 101
Generative AI products are built on one of two foundational architectures. Autoregressive models such as GPT, Claude, Gemini, and Llama generate text one token at a time, which is why latency and cost scale with response length. Diffusion models such as Midjourney, Stable Diffusion, and Sora start from random noise and iteratively denoise it into a coherent image or video. Modern systems combine these with multimodal capabilities by mapping text and images into a shared vector space. The model is only one piece; data pipelines, training methods, evaluation, safety, and serving infrastructure complete the system.
AI Concepts 101
Software 1.0 is hand-written logic. Software 2.0 shifted logic from code to curated datasets and neural networks. Software 3.0 turns natural language into the programming interface, making agents and vibe coding possible. In this new paradigm, a large language model becomes the central processing unit of an application, with context as its working memory, embeddings and vector databases as its persistent store, agent loops as its control flow, and tool calling as its hands. Evaluations replace traditional tests for non-deterministic behavior.
LLM Concepts, A Deep Dive
A large language model is an advanced autocomplete system. It predicts the next token in a sequence, one at a time, until it generates a stop signal. Its internal machinery has three layers worth knowing: tokens, embeddings, and parameters. Its training has three phases: pre-training on broad data, fine-tuning on a narrow dataset, and alignment through reinforcement learning from human feedback. Its interaction surface has two parts: the system prompt that shapes behavior and the user prompt that delivers the task. Its deployment has tradeoffs across proprietary, open-weight, and open-source variants, across small and large sizes, and across single-modality and multimodal designs.
Reinforcement Learning 101
Reinforcement learning is a branch of machine learning in which an agent interacts with an environment, takes actions, receives rewards or penalties, and gradually learns a policy that maximizes long-term reward. The vocabulary is compact: agent, environment, state, action, reward, policy, value, and episode. Modern large language models are refined through reinforcement learning from human feedback, which teaches them to prefer responses humans rate higher. Humanoid robots, game-playing systems that beat world champions, and production LLM assistants all share this learning loop.
How RAG Works
A base language model knows only what it was trained on. Retrieval-augmented generation solves that limitation by fetching relevant documents at query time and injecting them into the prompt before the model generates its answer. The system has two halves: an offline ingestion pipeline that chunks documents, embeds them, and stores them in a vector database, and an online retrieval pipeline that embeds the query, finds the most similar chunks, re-ranks them, and passes them to the model alongside the user's question. The model then generates a grounded answer with verifiable citations.
Context Engineering 101
Context engineering is the discipline of managing the information a model has at the moment it generates a response. Context includes the system prompt, conversation history, examples, tool outputs, retrieved documents, and any data the model can see. The context window is finite. Dumping more information into it often makes answers worse because models bias attention toward the start and end of the window. Good context engineering curates and composes the minimum useful information at each step, compresses what is stale, preserves what is durable in external notes, and splits work across sub-agents when a single context gets overwhelmed.
Context Engineering vs Prompt Engineering
Prompt engineering is the craft of writing a static instruction: role, examples, constraints, chain-of-thought triggers. It is useful, but it hits a ceiling the moment the model lacks information that was not in the prompt. Context engineering is the programmatic assembly of everything the model receives at inference time. It handles memory, retrieval, state, and tool access as a coordinated system. A prompt is a script handed to an improv actor. Context is the entire stage, props, and briefing dossier assembled before the actor speaks.
How AI Agents Work
A chatbot answers questions. An agent takes actions that have side effects. The difference is not intelligence, it is architecture. An agent combines a language model as the reasoning engine, a context window as its working memory, tools as its hands, and external memory as its long-term storage. Its control flow follows the ReAct loop: reason, act, observe, reason again, until the goal is reached. Plan-and-execute is a faster alternative when the path is known in advance. In practice, most systems blend both, adding human checkpoints before any irreversible action.
How to Design an AI Agent
A durable agent is the product of a disciplined evolution from manual work to autonomous operation. You start by doing the task yourself and documenting the runbook. You then use a language model as an assistant. You add tools through the Model Context Protocol so the agent can act on external systems. You teach the agent procedures through reusable skills. You give it memory. You orchestrate its work through standard operating procedures. You package it as a named agent. You schedule it. You integrate it with production triggers. Finally, you scale to multiple specialized agents coordinated by an orchestrator. At each stage, the human moves right along the workflow, doing less direct work but holding more accountability.
How MCP Works
Every AI integration used to be a custom build. With N AI applications and M data sources, that meant N times M bespoke connectors. The Model Context Protocol, released by Anthropic in late 2024, replaces that pattern with a single open specification. N clients plus M servers instead of N times M connectors. An AI host such as Claude Desktop or an IDE runs a client that talks to one or more servers, each of which exposes a database, an API, a file store, or an action. The protocol carries three primitives: resources that the model can read, prompts that are reusable templates, and tools that the model can call to take action.
How ChatGPT Apps Work
ChatGPT Apps replaced the earlier Plugins and GPTs approach with a richer architecture. Instead of returning plain text, an app can render a full interactive widget directly inside the conversation. The app has three parts: a Model Context Protocol server that exposes tools and UI resources, a widget that runs in a sandboxed iframe with its own state and event handling, and ChatGPT itself as the host that decides when to invoke the app and renders the widget. Users install apps through settings; conversations can then summon widgets automatically based on context or by explicit mention.
AI Coding Workflow 101
The workflow that produces reliable AI-assisted code is a loop, not a single prompt. Context comes first: project background, rules, relevant files. A plan comes before code. Code is generated in small, reviewable steps. Review happens with both human and AI-assisted tools. Tests are written alongside the change, not later. Debugging produces an explanation before it produces a fix. The loop repeats until the change is solid. The tools change every quarter. The method is what survives.