writing

►~ maverickstudios.net/writing

$ ls -la essays/

all published essays — Research Essays with Code

series 01

Beyond Tokens

Why AI needs world models to reason, plan, and act reliably. Five essays on the architectural shift from token prediction to state modeling. Each essay ships with a reference implementation on AWS.

substack ↗ github ↗

01 The Token Trap Why LLMs feel intelligent but don’t understand Jan 2026

02 Why World Models Change the Game From tokens to state Jan 2026

03 Why Planning Breaks Token-Based Systems When reasoning meets reality Feb 2026

04 Learning the World — Why VL-JEPA 2 Matters From observation to latent dynamics Feb 2026

05 From World Models to Working Systems Making intelligence executable Mar 2026

series 02

Executable World Models

A research framework for deterministic agent systems. Trajectories, not outputs. Each essay ships with a working GitHub release under MaverickHQ.

substack ↗ github ↗

01 Agents Can Plan The case for trajectory-first architecture Mar 2026

02 Evaluation is a Primitive, Not a Report Why structural verification must be built in Mar 2026

03 Agents Need Worlds Environment interaction as a first-class concern Apr 2026

04 The Architecture of Intelligent Systems How the layers fit together Apr 2026

05 Closing the Learning Loop Evidence policy feedback from experiment trajectories Apr 2026

06 Decisions That Don’t Disappear Placing Claude in the agent slot May 2026

07 Tools Return Results. Environments Change the World. The distinction that changes what agents can do May 2026

08 The Adversarial Quality Loop What changes when proposals are challenged before they become actions May 2026

09 Building the Player-Coach Loop How the PlayerAgent, CoachAgent, and constraint schema work together May 2026