Learning contents

Agentic AI v2

Agentic AI v2 is the next Kavriq learning path for building agent systems from first principles.

This section is currently being written. New articles will appear here as they are published, so treat this page as a live roadmap rather than a finished course.

If you want the complete set of articles today, use Agentic AI v1. It is slightly older, but it already covers the full foundations track and the production-oriented engineering series.

Why a v2?

The first Agentic AI path introduced the major concepts. v2 reorganizes the material around a sharper engineering question:

How do we build reliable systems around probabilistic reasoning?

Agentic AI is changing quickly. The vocabulary, engineering patterns, frameworks, safety practices, and production techniques are all moving fast enough that a full rewrite is more useful than slowly updating the existing articles in place.

That means more emphasis on state, feedback, tool contracts, checkpoints, recovery, evaluation, human boundaries, and production readiness. The goal is not just to explain agents, but to build durable intuition for designing them.

When v1 was written, the ecosystem was in an exploratory phase focused on defining basic primitives. Now that agentic development has matured, the focus must shift from what agents can do to how we build them reliably. v2 moves away from broad feature coverage—such as GUI automation or specialized computer vision—to center on core systems engineering. Regardless of what tools your agent uses, v2 teaches you how to design the control loops, state boundaries, and failure-recovery mechanisms around them.

Furthermore, because modern AI assistants now handle the majority of raw implementation code, v2 shifts focus from syntax-level boilerplate to architectural steering. Software engineering in an agent-driven ecosystem requires strong conceptual clarity: if you master the underlying mental models, you can direct, evaluate, and audit AI-generated systems to enforce robust architectural standards.

Once v2 is fully live, v1 will be phased out.

Scope & Depth

Agentic AI v2 is intentionally more detailed and conceptually comprehensive than v1. Rather than providing quick code snippets or surface-level summaries, these 44 articles are designed to give you a complete, end-to-end mental model. The goal is to build deep architectural intuition so you can confidently design, audit, and debug agentic systems in production.

Current Status

Start Agentic AI v2 Foundations
The v2 foundations section is live and will expand article by article.
Read Agentic AI v1
Use this if you want the complete, slightly outdated version of the course now.

What v2 Will Cover

Agentic AI v2 is planned as a full systems-level curriculum: from the smallest definition of agency to production reliability, security, observability, deployment, and a capstone agent.