Aditya Karnam · World Model Infrastructure Lab

Field Notes / World Model Infrastructure Lab

Thought leadership notes for the systems layer behind AI agents.

This section reframes the site around the infrastructure needed for agents that maintain state, retrieve memory, route across models, simulate outcomes, and act with more reliability over time.Some of these essays already exist indirectly in project write-ups and technical notes. Others are planned field notes that make the research agenda explicit.

Research Scope

state + memory
retrieval + context
simulation loops
model routing
local inference
observability + evals

Published signals3

Priority essays3

Research threads6

ModePlanned + grounded

Publishing Thesis

What this section is for

Field Notes replaces a generic blog framing. The goal is to make the site read like an engineering research notebook rather than a list of disconnected posts.

Thesis

The next frontier of AI is not conversation alone. It is systems that can model the world: maintain state, remember context, route choices, and act through tools with less brittleness.

Why now

Bigger base models keep improving, but the missing layer is increasingly obvious when teams try to ship persistent, multi-step, tool-using agents in practice.

Editorial bar

Each note should make one strong claim, tie it to system design, and point back to public artifacts or experiments already visible on the site.

Grounded Signals

Existing posts that already support the story

These are the current notes and project write-ups that map most directly to the world model infrastructure framing.

July 01, 2026

subagent-fleet: Local AI Compute Control Plane for Coding Agents

I built subagent-fleet to route Claude Code-style subagents across local Ollama machines, with LiteLLM generation, health checks, model warmup, and a live dashboard.Strongest public proof of the runtime, routing, and local-compute thesis.Open note

April 05, 2026

embenx Guide: The Ultimate Python Library for Vector Search

Learn how embenx provides a unified API for 15+ vector backends like FAISS and pgvector, featuring temporal memory for 71% better retrieval performance.Best current artifact for memory, retrieval abstraction, and MCP-facing context systems.Open note

April 11, 2026

AI Blog Generator: I Built an n8n Bot That Wrote 139 Posts

I built an n8n workflow pulling from 7 RSS feeds, used Google Gemini to write MDX posts, and auto-published 139 blog posts to GitHub. Here's what Google Search Console showed.Shows workflow automation instincts and a willingness to instrument outcomes instead of hand-waving them.Open note

Planned Essays

The first field notes to publish

The first three are the core essays. The rest are the adjacent notes that help complete the research program.

Priority draft

The Missing Infrastructure Layer for World-Model AI

Foundation models are not enough for real-world agency. The next category is the infrastructure around them: state, memory, simulation, routing, and evaluation.Linked signals: subagent-fleet / embenx

Priority draft

From RAG to State: Why Agent Memory Is Not Just Retrieval

Retrieval gets facts back. Memory systems need to track evolving goals, tool use, environment state, and the consequences of prior actions.Linked signals: embenx / awesome-agentic-memory

Priority draft

Local-First AI Infrastructure for Agent Builders

As agent workflows get longer and more expensive, local inference, model routing, and hybrid compute become infrastructure advantages rather than hobbies.Linked signals: subagent-fleet / MLX non-determinism

Planned field note

Why LLM Agents Need State

Stateless prompting breaks down once tasks span time, tools, retries, and user-specific context.Linked signals: subagent-fleet / awesome-agentic-memory

Planned field note

Why Model Routing Matters for Agentic Systems

Routing is where cost, latency, capability, locality, and reliability meet. One model endpoint is not a systems strategy.Linked signals: subagent-fleet / AI Toolkit

Planned field note

World Models Will Need Observability

If agents simulate and act, builders need traces that expose why they chose a route, tool, or memory update.Linked signals: subagent-fleet / MLX non-determinism

Reading Paths

How to navigate the material

Static reading paths keep the section useful before any dynamic AI layer or RAG interface is added.

World models

Start with the infrastructure thesis below.
Read subagent-fleet for the runtime and routing angle.
Follow with embenx for retrieval and memory interfaces.
Treat the remaining essays as the planned research map.

Local inference

Begin with subagent-fleet for fleet orchestration and model routing.
Use MLX non-determinism as the reliability counterweight.
Then connect both to the planned local-first infrastructure essay.

Next moves

The highest-leverage publication sequence is the infrastructure thesis, then memory/state, then local-first routing. That creates a coherent narrative arc before expanding into observability, MCP, or evaluation.

Read the positioning Start with subagent-fleet