MotyWait field notes · AI engineering
Concise engineering references connected to Mira, the course labs, and one complete path from model output to a production-ready agent.
Eight connected modules
Each module connects the teaching deck to a compact reference and a concrete lab. The complete 493-slide instructor deck and answer keys remain private while the course is being completed.
Understand probabilistic output, tokens, context limits, and the five-part prompt structure used throughout the course.
Mira can communicate with a defined role, task, constraints, and output shape.Compare prompt structures and convert an unstructured response into a validated result.
Move from a single model response to a controlled loop that can select tools, act, observe results, and stop safely.
Mira can choose a permitted tool and use its result without seeing its credentials.Design a minimal tool set, schemas, permission boundaries, and failure behavior.
Read the related guide ↗Compose routing, chaining, decomposition, and evaluator patterns into a small agent that is easy to inspect.
Mira can break a goal into bounded steps and explain what happened.Build the travel-agent core, then test tool errors, incomplete inputs, and unsupported requests.
Separate working context, session state, durable preferences, and source-of-truth data instead of calling everything memory.
Mira can remember an approved preference without leaking it into another user’s session.Add explicit save/retrieve behavior and verify isolation across sessions and users.
Read the related guide ↗Build ingestion, chunking, embeddings, retrieval, grounding, and citations as one measurable system.
Mira can answer from approved documents, cite evidence, and abstain when evidence is weak.Create a small knowledge base, inspect retrieved chunks, and test unanswerable questions.
Read the related guide ↗Treat cost, latency, reliability, observability, and deployment as architecture—not cleanup after the demo.
Mira can operate within a latency target, cost budget, and defined fallback path.Measure a baseline, remove unnecessary context, add caching, and compare before/after behavior.
Test task quality and system safety with datasets, thresholds, permissions, adversarial cases, and human escalation.
Mira can refuse unsafe actions, protect secrets, and surface uncertain decisions to a person.Create quality, injection, exfiltration, excessive-agency, and escalation tests.
Combine goals, tools, memory, knowledge, guardrails, evaluation, and operations into one defensible system.
Mira can complete a bounded workflow and show the evidence needed to review it.Deliver an architecture brief, working flow, evaluation results, threat review, and operating plan.
Start with a question
02 · Agents & tool use
Understand the controlled loop, tools, state, permissions, and stop conditions that distinguish an AI agent from a chatbot.
Read the guide ↗05 · Knowledge & retrieval
A practical map of ingestion, chunking, embeddings, retrieval, citations, abstention, and evaluation in a production RAG system.
Read the guide ↗04 · Memory & context
Separate context, session state, durable preferences, and organizational knowledge to build safer, more useful AI-agent memory.
Read the guide ↗Course 01 · development list
Join the AI Engineering list for field notes, sample material, and the release announcement. No enrollment or payment is open yet.
