68 topics across 10 stages, in the order that actually builds on itself — from what a token is, through prompting and context, to RAG, agents, MCP and evaluation. Written for people starting from zero.
Each stage assumes the one before it. Start at 00 if you are new, or jump to the first thing on the map you could not explain to someone else.
What the role is, and a workspace you will keep.
Tokens, context windows, hallucination, sampling.
System prompts, chaining, structured output.
Compaction, memory, and long-session failure modes.
Vectors, chunking, and semantic search.
Give a model your data without retraining it.
Tool loops, memory, and multi-agent systems.
One standard way to plug tools into any model.
When to change the model itself — and when not to.
Scoreboards, tracing, cost and hallucination checks.
Tutorials publish Monday, Wednesday and Thursday. Fridays are a hands-on project that ties the week together.
Long AI conversations drift. Agent memory leaks. Embedding models vary wildly in quality. This advanced project ties all three problems together into one monitoring dashboard — built entirely by prompting Claude or ChatGPT. You'll walk away with a real tool and a much deeper understanding of how AI systems actually manage knowledge over time.
Choosing an embedding model is one of those RAG decisions that looks simple on the surface but quietly shapes your retrieval quality, storage costs, and monthly bills. This article breaks down the three tradeoffs that actually matter — dimensions, cost, and quality — and walks you through a practical comparison framework you can run on your own data in under an hour.
AI agents need to remember what they've done, what they've found, and where they're going — but carrying too much context kills performance. This guide breaks down the four things an agent needs to remember, where to store state, and practical patterns to stop your context from ballooning out of control.
Reading only gets you so far. Each project is a step-by-step guide to prompting an AI tool into building something real.
Prompt AI to build a secure API key manager from scratch with step-by-step instructions.
Start buildingBuild a personal learning tracker with progress bars, streaks, and data visualization.
Start buildingUse AI to create a searchable directory of 20+ AI plugins and MCP servers with filtering and favorites.
Start buildingPrompt AI to build an interactive quiz that recommends Cursor, Copilot, or Claude Code based on preferences.
Start buildingBuild a reusable prompt template system with variable placeholders and custom creation.
Start buildingPrompt AI to build a multi-step wizard that generates valid MCP server configurations.
Start buildingNew tutorials three times a week, working through the roadmap in order.
No email required — subscribe with any RSS reader and they arrive as they publish.
https://danbyers.ai/feed.xml