top of page

Agentic AI for CNO Development

Course Description

​

AI has drastically changed CNO. Well-established roles are merging or being replaced, and tactics are evolving at a pace that no one was prepared for. We need to reorganize and rearm because AI is not going anywhere.

​

In this course, we cover the current AI ecosystem and ways to use it. The examples, demos, and exercises are all focused on ways to leverage AI in the CNO lifecycle — from mission planning to initial access, persistence, operations, analysis, and reporting.

​

This course is hands-on and driven by practical exercises, demos, and discussion. Students use a preconfigured VM with all required tools, frontier AI models, and a GitLab server to version control all materials.

​

Learning Objectives:

1. How to use AI tools, agents, agent teams, and full orchestration

2. Understand strengths and weaknesses of AI and when to use it

3. How to apply AI throughout the CNO lifecycle

​

Prerequisites: Experience in CNO and programming in C, Python, and Linux.

Class Format

​

Hands-on keyboard. Students receive a preconfigured VM with all required tools, IDEs, and utilities. All exercises, demos, and course materials are accessible via GitLab.

​​

Topic Details

Day 1: AI Ecosystem

​

1. AI in CNO development: from planning to development and maintenance 

2. LLMs, tools, agents, agent teams, and full orchestration 

3. Human in the loop, on the loop, and out of the loop 

4. Shifting roles in development, reversing, development, devops 

5. Generate images, videos, text, and code; voice cloning 

6. Git version control at the speed of AI development 

7. Risks of underestimating and overestimating AI 

8. LLM lifecycle: training, fine-tuning, calibrating, restricting, and inference 

Day 2: Agents

​

  1. Agent hooks, skills, subagents, plugins, headless mode 

  2. Session management: forks, saves, resumes 

  3. AI model context windows, tokens, and quantization 

  4. Types of prompts: system, user-wide, and project prompts 

  5. Prompt engineering: jailbreaking, meta-prompting 

  6. Workflow changes: snapshots, version control worktrees, guard rails 

  7. Knowledge management: progress reports, 

Day 3: Agent Teams

​

1. Scaling tasks in depth and breadth, parallel tasking 

2. Terminal multiplexing for agent management 

3. Several agents parallelize the development lifecycle 

4. Interfacing with AI as work roles: junior, peer, senior, and project manager 

5. Local vs remote LLMs, privacy and information leaks 

6. Limitations and risks: cutoff dates, hallucinations, context window 

7. Practically integrating AI into live operations 

8. Company specifics: Anthropic, OpenAI, Google, xAI, Meta, Amazon 

Day 4: Full Orchestration

​

  1. Virtual machines, containers, and devops for AI 

  2. CI/CD pipeline integration 

  3. Integrating online with offline models: power, privacy, and probability 

  4. Weaning: migration from remote and probabilistic to local and deterministic 

  5. Self-hosted AI: Ollama, Hugging Face, OpenWebUI, LiteLLM 

  6. Model agnostic AI IDEs and CLIs: Kilo, Cline, RooCode, VS Code 

  7. Building harnesses and feedback loops 

  8. Retrieval augmented generation (RAG) 

Day 5: Capstone Project

​

1. The best way to learn AI is to tackle a big, relevant task 

2. Students apply all lessons to do an end-to-end project of their choice 

3. Generate requirements and corresponding plan 

4. Design testing criteria and employ testing framework 

5. Leverage AI to reach requirements 

6. Document status and share 

7. Path forward and how to stay current: innovators, blogs, Google Alerts 

8. Open discussion 

bottom of page