Research Articles & Technical Deep Dives

Infrequent, high-signal research on AI architecture, agentic systems, cognitive architectures, and systems design by Aditya Gaurav.

Shared caches beat sandbox walls

SEP 14, 2026 · Architecture

In July 2026, OpenAI agents in ExploitGym runs turned an internal Artifactory cache into a message board, coordinated across supposed isolation boundaries, and breached Hugging Face while chasing a…

Read full article →
The Chat Template Trap: Why Dynamic Agent State Crashes Open-Weight Models

SEP 02, 2026 · Architecture

Multi-agent systems cannot run on static prompts. But the moment you inject volatile state into open-weight inference, Jinja templates throw fatal exceptions. Here is how we engineered asymmetric…

Read full article →
Memory Stacks for Agents: A Deep Dive into Episodic, Semantic, and Procedural Architecture

APR 16, 2026 · Framework

Memory is the layer below reasoning, and most agent teams build it wrong. A systems architect's field guide to the three kinds of memory your agent actually needs, and how to make them work together.

Read full article →
Inference Economics for Agents: A Research Framework for Why Token Cost Is the Wrong Metric

APR 13, 2026 · Framework

Every team building agents in 2026 is optimizing the wrong number. Token cost tells you a fraction of the story. Here is a framework for the other four fractions.

Read full article →
Software 3.0: What It Actually Means to Architect AI Systems

JAN 15, 2026 · Architecture

This framework is elegant and widely cited. But it stops at the naming layer. It tells you what the eras are. It doesn't tell you what it means to be a systems architect in Software 3.0.

Read full article →
The "Product Is the Model" Principle - And How to Actually Apply It

NOV 10, 2025 · Principle

"The product is the model. The approach is to expose the model as directly as possible, with minimal scaffolding, and allow the model to determine the best approach."

Read full article →
The Verification Gap: Why Reasoning Models Need Deterministic Arbiters

SEP 02, 2026 · Principle

Test-time compute and extended chain-of-thought are not magic. Beyond a critical token budget, unconstrained reasoning leads to prior drift, verifier gaming, and overthinking. The governing…

Read full article →
The Invariant Problem in Autonomous Remediation: Why Code Synthesis Fails Without Closed-Loop Verification

AUG 28, 2026 · Framework

Generating a patch that silences a crash is easy. Guaranteeing that the patch does not break downstream invariants, truncate valid protocols, or leak resources is hard. Here is an architectural…

Read full article →
Sandboxing Architectures for Autonomous Agents: MicroVMs, gVisor, and the Egress Dilemma

AUG 14, 2026 · Architecture

Running autonomous coding and offensive security agents inside standard Docker containers is an operational hazard. Here is a systems guide to isolation runtimes, cold boot mechanics, and the…

Read full article →
Context Circuit Breakers: Preventing Reasoning Collapse in Long-Horizon Agent Loops

JUL 28, 2026 · Framework

When autonomous agents run past 20 turns, standard context windows do not just fill up. They degrade. Here is an architectural framework for deterministic circuit breakers, tool output quarantine,…

Read full article →
The MCP Security Paradox: How the Protocol That Connects Everything Also Exposes Everything

MAY 01, 2026 · Framework

A critical vulnerability in the Model Context Protocol surfaced in April - 150 million downloads exposed to remote execution risk. Cloudflare published a reference architecture. AWS Bedrock…

Read full article →
The Big AI Systems Architecture Comparison

MAR 05, 2026 · Architecture

Raschka compared LLM architectures across 7 years. This compares AI systems paradigms across 5 years of production deployment.

Read full article →
The 6 Paradigm Shifts Nobody Told You About (2026 Edition)

JAN 02, 2026 · Architecture

An AI Systems Architect's Year-in-Review - Inspired by Karpathy's format, written for the systems layer.

Read full article →