Version: 0.3.0 · Changelog · SemVer
Free public course by iMPLEMENTAi.ae · Course landing page
A practical, build-first course for learning how to operate, control, build, test and improve useful AI agents and agentic systems.
This repository is the canonical general course. Individual learner repositories may adapt examples, pace and domain projects while preserving the core learning outcomes.
Start your first lesson with David AI Coach — copy one self-contained prompt into ChatGPT (or your AI chat app), receive a worked example and complete your first exercise. You do not need GitHub registration, GitHub access, connectors, a terminal, downloads or file uploads. GitHub is taught in Module 4; it is not a setup gate.
Open the guided curriculum and roadmap — 17 ordered modules, practical outcomes and estimated time for each, followed by a graduation challenge.
Plan for 50–75 active learning hours: roughly 10–15 weeks at 5 hours/week or 5–8 weeks at 10 hours/week. These are adjustable planning estimates. Each session shows where you are, what you will learn and what comes next.
Start with Module 0 — Welcome, then Module 1 — AI mental models. David AI Coach uses supplied background and teaches immediately; there is no baseline questionnaire to complete before you can start.
David AI Coach is your AI tutor, guided by David Potgieter's practical teaching method: explain, build, inspect, debug, verify and teach back. It adapts the course to your demonstrated ability and resumes from the learning evidence you save.
Use your own ChatGPT, Codex, Claude or another capable AI environment. The course instructions activate the coach there. The curriculum is free; your chosen AI provider's usage limits and charges still apply.
Setup and installation guide — start in a browser or download the course for hands-on work.
Use Start first lesson. The coach can teach Module 0 and the first part of Module 1 from that self-contained prompt even if your AI cannot read this repository. No manual folder navigation or Markdown uploading is required to begin. For optional later technical steps, see Setup.
The course is based on operating patterns proven in real AI-agent work:
- one capable agent before multi-agent complexity;
- project instructions before giant one-off prompts;
- GitHub as durable working memory and change evidence;
- Skills for reusable HOW;
- SB1-style smart-agent-first Skill design;
- KA1 / KISSS simplification before adding machinery;
- Lessons that change future behaviour, not lesson logs for their own sake;
- one canonical owner/source of truth;
- deterministic helpers only when they provide real leverage;
- verification as part of execution;
- real security boundaries without approval theatre.
See METHODOLOGY.md.
AI mental models
→ operate one agent well
→ project instructions / AGENTS.md
→ files + Git + GitHub
→ reusable Skills
→ Codex and coding literacy
→ tools + APIs + MCP
→ context + state + memory + data
→ debugging + evals + safety
→ smallest useful architecture
→ real capstone agent
→ multi-agent systems
→ production thinking
See COURSE.md for the complete roadmap.
Final proof: GRADUATION.md · Continued practice: CONTINUATION.md · Current harness practice: Pi + DeepSeek Harness
- AGENTS.md — operating instructions for the learner's AI coach
- COURSE.md — full curriculum
- METHODOLOGY.md — the operating principles behind the course
- course/ — core lessons and graduation work
- labs/ — hands-on technical labs
- workshops/ — guided builds
- exercises/ — domain practice
- prompts/ — reusable prompts
- playbooks/ — practical operating/debugging procedures
- knowledge/ — reference knowledge
- templates/ — small starter templates
- rubrics/ — evidence-based verification criteria
- checklists/ — proportional production checks
- tests/ — cold-start/course-operability tests
- work/ — learner working area
- outputs/ — completed artifacts
- .folderdesk/skills/ — small FD Tiny-derived local Skill foundation
The canonical course stays general.
A learner-specific repository may add:
- personal background and technical level;
- preferred analogies;
- domain projects;
- pace adjustments;
- personal exercises;
- capstone specialization.
Example: Course-Andre-Venter.
Personalization must not become a competing source of truth for the general curriculum.
This course openly publishes the operating principles we use at iMPLEMENTAi.ae when turning capable AI into useful business capability.
The course stays vendor-light at the principle level, but it also includes a maintained current practice layer so learners can work with useful tools of the moment. As of September 2026, that includes Pi and DeepSeek Harness for hands-on harness and multi-agent experimentation.
The same principles also sit underneath FolderDesk — the deeper Business AI Operating System we use when a business needs durable context, reusable Skills, connected systems, operating history and agent work that must continue beyond individual chats.
The course is free. The commercial work is optional.
Learn by operating and building.
A concept should become an action, file, diff, Skill, test, tool call, diagnosis or verified result as quickly as practical.
The course is complete when the learner can independently take a real problem and build a reliable agentic solution—not when every page has merely been read.
Use the course yourself first. That is what it is here for.
If you want hands-on help turning the same principles into working AI capability around your actual business:
- AI Build With You — work directly with David on real AI opportunities, Skills, workflows, context and useful system connections. Current engagement: AED 3,700/month.
- Free AI Implementation Session — bring one real bottleneck and use 30 minutes to identify and, where practical, start implementing the most useful AI opportunity.
- FolderDesk — explore the deeper operating layer when AI work needs durable context, reusable capability and continuity across tasks, systems and people.
Start free: Course Agentic AI · Open the public GitHub course