01 / Evidence before confidence
Separate a generated answer from verified evidence. Define success, retain sources and identify uncertainty.
LEARNING PATHS / A SYLLABUS WITH INTENTION
Choose a destination.
Give the learning a route.
A pathway is a sequence of questions, exercises and evidence. Adapt the pace to your background; move on when you can explain and demonstrate the outcome.
YOUR SUGGESTED ROUTE
Start with an outcome.
- Choose a goal.
- Set your starting point.
- Build a route you can adapt.
THE SHARED CORE
Four classes.
A foundation you can reuse.
02 / Read the workflow
Follow inputs through a graph, preserve settings and change one variable in a controlled comparison.
03 / Design the environment
Understand memory, dependencies, data location, access and the need for a tested fallback.
04 / Bound the agent
Define the task, permitted tools, approval points, stop conditions and evaluation before adding autonomy.
PRACTICE BY DESTINATION
Make a small portfolio
of things you can explain.
Creative images
Map a simple graph → generate a baseline in your chosen tool → vary one setting → compare composition and artefacts → save a reproducible project record.
Local AI systems
Define the workload → inventory available hardware and software → choose a supported installation route → measure a small run → document resource limits and fallback.
Agentic workflows
Map a manual task → define an agent role → specify a read-only tool contract → test missing or hostile context → review a draft → permit further action only within agreed authority.
Voice and video
Write a short media brief → use owned or licensed inputs → generate a small sample in a chosen service → check intelligibility, continuity and identity permissions → document edits and provenance.