Coding with Agents

Appearance

Level 5: Be an AI Architect ​

Last updated: 2026-08-06

The tools will keep changing. The mindset you build now determines whether you lead the change or react to it.

The first four chapters equipped you with progressively larger scopes: prompting (individual communication), anatomy (understanding the machine), power-user techniques (individual mastery), and team practices (scaling across people). This final chapter completes the arc by stepping back from the keyboard entirely. Here, we ask the bigger questions.

Reading Order ​

  1. Product Management in the AI Era — production shifts from an assembly line to overlapping roles (vibe working), and the PM becomes an AI builder. Once anyone can build the thing, survival moves to four stages outside the product: arena, traffic, retention, revenue.

The remaining three pieces are outlined below and are not yet written; none are in the navigation yet. Progress is tracked in ROADMAP.md.

  1. The paradigm — Software 3.0 will introduce Karpathy's three-generation model of software: hand-written code (1.0), learned neural networks (2.0), and LLMs as programmable reasoning engines (3.0). Understanding this framing helps you see where today's tools fit in the larger arc of computing — and why prompt engineering and context management are genuinely new forms of programming, not just fads.

  2. The mindset — Builder's Mindset argues that in a Software 3.0 world, engineers must become architects of AI-powered systems rather than solely writers of explicit code. The messy, interconnected reality of real software systems demands people who can think in systems, manage ambiguity, and direct AI toward the right problems.

  3. The trajectory — The Future of Software Engineering will map where the industry is heading: how developer roles are changing, what AI-native workflows look like, and what skills will matter most in the years ahead.

Where you've been: From your first prompt in Chapter 1 to team-scale practices in Chapter 4, you've built a complete toolkit for working with AI coding agents. This chapter is the lens that ties it all together — the "why" behind the "how."

The invitation: The developers who will thrive aren't the ones who ignore these tools or the ones who blindly trust them. They're the ones who understand the tools deeply enough to direct them effectively — and who never stop asking what should be built, not just what can be built.