Building Agentic AI Workloads – Crash Course
•January 6, 2026
0
Why It Matters
Understanding agentic AI is essential for businesses to leverage emerging autonomous capabilities while managing rapid technological change and associated risks.
Key Takeaways
- •Generative AI scales data, models, compute dramatically across industries
- •Agents add autonomy beyond simple LLM output generation
- •Parallelized transformers enabled training of trillion‑parameter models at scale
- •Model‑as‑a‑service shifts AI development to large providers globally
- •Rapid evolution demands attention to timestamps and emerging standards
Summary
The video provides a crash‑course on agentic AI, tracing the evolution from early AI research to today’s generative‑AI boom and the emergence of autonomous agents.
Raali explains that AI rests on three pillars—algorithms, data, compute—and shows how each has exploded: datasets now span terabytes to petabytes, models have grown from millions to trillions of parameters, and parallelized transformer architectures have turned serial training into massive parallel workloads. This convergence created “Hulk” foundational models that can understand and generate human language across a wide range of tasks.
Key milestones are highlighted: the 2017 transformer paper, OpenAI’s ChatGPT launch in 2022, and the 2023 ReAct paper that merged reasoning with action, ushering in the “year of agents.” Raali notes that agents now control file systems, browsers, and can spawn other agents, while the term “agentic systems” is used to capture the spectrum of autonomy.
For enterprises, the shift to model‑as‑a‑service means AI capabilities are now offered by a few cloud giants, making in‑house development cost‑prohibitive but also opening new integration opportunities. Because the field evolves faster than academic consensus, practitioners must track timestamps on research and be cautious about deployment risks such as hallucinations and shifting best practices.
Original Description
This course, from Rola Dali, PhD, provides a comprehensive overview of agentic AI, defining agents as software entities that use LLMs to perceive environments, make decisions, and execute actions to achieve specific goals. It explores the critical distinction between static workflows and dynamic agentic systems, emphasizing how LLMs serve as a reasoning "brain" to decompose tasks at runtime. Through practical Python demonstrations, the course covers essential components like system prompts, tools, and memory, while also comparing architectural patterns such as Supervisor and Swarm. Finally, the session addresses the future of technology by discussing emerging interoperability protocols like MCP and the shifting paradigms of software development in an AI-driven world.
Slides and Labs: https://github.com/rdali/ML105_Agents
❤️ Support for this channel comes from our friends at Scrimba – the coding platform that's reinvented interactive learning: https://scrimba.com/freecodecamp
⭐️ Contents ⭐️
- 0:00:00 Introduction and Speaker Background
- 0:01:15 A Brief History of Artificial Intelligence (1940s–Present)
- 0:05:43 Traditional Machine Learning vs. Generative AI
- 0:06:35 The Three Pillars of AI: Algorithms, Data, and Compute
- 0:11:08 Specific Tasks vs. General Task Execution
- 0:14:41 Defining Agency and the Spectrum of Autonomy
- 0:18:00 Agentic Milestone Timeline (2017–2026)
- 0:20:31 What is a Generative AI Agent?
- 0:23:04 Agents vs. Workflows: Dynamic Flow vs. Static Paths
- 0:26:18 Pros and Cons of Agentic Systems
- 0:29:59 Patterns and Anti-patterns: When to Use Agents
- 0:32:36 The Core Components of an Agent
- 0:34:55 Choosing the Right LLM for Your Agent
- 0:37:38 Crafting Identity with System Prompts
- 0:39:00 Understanding Memory: Intrinsic, Short-term, and Long-term
- 0:41:26 Enhancing Capabilities with Tools and Actions
- 0:43:09 Hands-on Implementation: From Single LLM Call to Python Agent
- 0:52:18 Adding Memory and History to Your Custom Agent
- 0:54:53 Building Agents with Frameworks (LangChain)
- 0:57:17 The Evolving Landscape of Models and Frameworks
- 1:00:15 Agentic Architectural Patterns: Supervisor vs. Swarm
- 1:01:41 Case Study: Single Agent vs. Supervisor Architecture
- 1:04:48 Deep Dive: Swarm Architecture Performance
- 1:06:08 When to Choose Multi-agent Systems
- 1:09:05 Interface Protocols: MCP, A2A, and AGUI
- 1:12:06 How to Evaluate Agentic Systems (LLM vs. System vs. App)
- 1:13:53 Evaluation Methods: Code-based, LLM-as-a-Judge, and Human
- 1:15:25 Current Challenges: Hallucinations, Cost, and Debugging
- 1:18:15 Real-world Incidents and the AI Incident Database
- 1:21:28 Career Impact: Which Jobs are Most at Risk?
- 1:23:41 Software 3.0: The Evolution of Development Paradigms
- 1:29:00 Weathering the Storm: Strategies for the Future
- 1:33:40 Beyond LLMs: World Models and the Future of AMI
- 1:37:15 Recommended Resources and Closing Thoughts
🎉 Thanks to our Champion and Sponsor supporters:
👾 @omerhattapoglu1158
👾 @goddardtan
👾 @akihayashi6629
👾 @kikilogsin
👾 @anthonycampbell2148
👾 @tobymiller7790
👾 @rajibdassharma497
👾 @CloudVirtualizationEnthusiast
👾 @adilsoncarlosvianacarlos
👾 @martinmacchia1564
👾 @ulisesmoralez4160
👾 @_Oscar_
👾 @jedi-or-sith2728
👾 @justinhual1290
--
Learn to code for free and get a developer job: https://www.freecodecamp.org
Read hundreds of articles on programming: https://freecodecamp.org/news
0
Comments
Want to join the conversation?
Loading comments...