Six Waves of AI Coding

A framework for understanding the rapid evolution of AI-assisted software development and its career implications.

Core Thesis

AI-powered coding agents represent a paradigm shift that will reshape software development within months, not years. Each wave arrives faster than its predecessor, compressing adoption timelines.

“It’s not AI’s job to prove it’s better than you. It’s your job to get better using AI.”

The Six Waves

Each wave provides roughly 5x productivity gains over the previous:

Wave 3 → 4: Chat to Agents

The critical transition happening now:

Chat coding:

• Continuous back-and-forth dialogue
• Developer drives the conversation
• Manual iteration and correction

Coding agents:

• Operate autonomously in terminal environments
• Handle multi-step tasks independently
• Only require intervention when stuck

Key insight: “Vibe coding” (letting AI handle code writing) is a mindset that works across all modalities, not tied to any single tool.

Wave 5-6: The Fleet Model

Agent Clusters (Wave 5):

• Multiple agents working in parallel
• Separate git worktrees for isolation
• Human coordinates across agents

Agent Fleets (Wave 6):

• Supervisor agents manage groups of coding agents
• Humans transition to “fleet management”
• Each IC effectively manages a hierarchical AI workforce

The organizational structure shifts from:

Human → Code

To:

Human → Supervisor Agent → Coding Agents → Code

Economic Reality

Companies unable to fund this infrastructure face competitive disadvantage. Agent clusters necessitate cloud-based development environments.

Career Implications

The Inversion Dynamic

• Junior developers adopt AI tools enthusiastically with minimal ego resistance
• Senior developers often resist due to identity tied to craft mastery
• Economic pressure forces difficult budget decisions

This creates a potential inversion where experience becomes a liability when paired with resistance to change.

Skill Shift

The defining skills change from:

• Writing code efficiently
• Deep language/framework expertise
• Manual debugging prowess

To:

• Task decomposition for agent delegation
• Agent supervision and course correction
• Quality verification of AI output
• Fleet coordination and prioritization

Task Decomposition

Coding agents require careful scoping:

Too large: Agents fail spectacularly, lose context, spin on errors

Right-sized: 3-20 step workflows with clear boundaries

“Oversizing tasks causes agents to fail spectacularly.”

This aligns with Factor 10: Small, Focused Agents.

Timeline Pressure

The compressed timeline means waiting to adapt is increasingly costly.

Strategic Takeaways

For individuals:

1. Immediately experiment with coding agents
2. Transition from completions to chat-based development now
3. Learn agent management before the cluster wave arrives
4. Adaptability and AI fluency trump years of experience

For organizations:

1. Budget for significant LLM expenditure increases
2. Plan for cloud-based development infrastructure
3. Evaluate hiring for AI-native developers
4. Early adoption creates compounding advantage

Key Principle

“The speed of AI improvement means resistance is career-limiting. The developers who thrive will be those who learn to multiply their effectiveness through AI, not compete against it.”

Related

• Skill Atrophy – What skills to maintain vs. delegate
• Highest Leverage: Plans & Validation – Where humans add most value
• The Meta-Engineer Identity – Building systems that build systems
• Agent Swarm Patterns – Parallel agent coordination