Series
The Eightfold Path of AI
A mental model for understanding why AI works, why it changed so suddenly, and where it is heading next. Eight steps that build on one another - from the earliest computational constraints to systems that initiate their own work.
The Historical Constraint
Pre-AI
For decades, limited computation defined what AI could be. Why the ideas existed long before the hardware could run them.
From Rules to Learning
Data and Pattern Recognition
The shift from hand-crafted rules to statistical learning - where computers stopped being told what to expect and started finding structure themselves.
Hierarchies and Abstraction
Deep Learning and Representation
How neural networks began discovering hierarchies and abstractions that no human explicitly specified - the end of hand-engineered features.
The Architecture of Attention
The Transformer
The 2017 breakthrough that changed everything. Instead of processing sequences step by step, models learned what to pay attention to.
General Purpose Learning
Foundation Models
Scale changed the game. One engine to support many applications - GPT-3, DALL-E, and the arrival of AI as a general platform.
Toward Autonomous Capability
Agentic Systems
The shift from reactive to directional - systems capable of planning, using tools, and managing tasks that unfold over time.
Controlled Action
Orchestration
Agents become systems. Explicit workflow, state management, and execution boundaries - the scaffolding that keeps AI safe and operational.
When Systems Start Initiating Work
Agentic AI
The final shift: from orchestration to autonomy. Systems that pursue goals with limited supervision - and the governance questions that follow.