Control Systems × AI

Why Current LLM Architectures Behave Like Open-Loop Systems

And why stability requires a control-theoretic layer.

open-loop vs closed-loop · feedback control · runtime stability · autonomous correction

Why LLMs Fail in the Real World: Design-Time Optimization vs Real-Time Stability

Why modern LLM systems overfit the setup phase and collapse during execution.

preparation vs survival · fine-tuning limits · control-theoretic reframe

Hierarchical Control for Browser Automation

Applying supervisory control and distributed systems theory to agentic architectures.

state-space decomposition · supervisory control · sensor fusion · DMPC · failure isolation

Designing systems in non-deterministic environments.

stability · convergence · dynamic systems · feedback loops · architectural resilience

Why architecture, not models, determines long-term product advantage.

architecture · constraints · invariants · stability · competitive moats · control over intelligence

A foundational primitive for distributed browser intelligence.

browser contexts · session sharing · parallel pages · sub-agent delegation · createPage()

Legacy Invariants: One Quiet Reason AI-Driven Automation Breaks Unexpectedly

Hidden UI assumptions from early 2010s web frameworks.

timing constraints · human-like interactions · legacy systems · invisible assumptions

Enterprise Application Models

Where theory becomes production systems.

HR · payroll · compliance · onboarding · document processing · browser operations

Failure Paths & Correction Loops

Deterministic recovery without drift.

structured failure semantics · fallback operators · safe recovery · no blind retries

Operator Composition (Pipelines & Multi-Step Graphs)

Composable operators, not monolithic agents.

verified pipelines · DAGs · localized failures · scalable workflows

Deterministic Planning via Structural Constraints

Structure is the reasoning.

structural constraints · action primitives · workflow graphs · automation safety

Zero‑Context Architecture

The simplest path to reliable agents.

domain isolation · no history · reproducibility · enterprise determinism

Domain Extractors & Problem Rewriting

Converting messy inputs into solvable representations.

canonical forms · entropy reduction · manifold contraction · 0-context foundations

Closed-Loop Verification Architecture

Proposals meet guarantees.

deterministic validation · error boundaries · stable transitions · enterprise reliability

State-Space Model for Agentic Systems

Mathematical structure for reliable agents.

discrete-time dynamics · domain projection · verification · error dynamics

Agents as Operators, Not Oracles

Transformation, not reasoning.

control theory · deterministic transformation · verifiable systems · operator pipelines

Domain Mismatch & Hallucination

Correct operators on incorrect representations.

attractor misalignment · structural ambiguity · domain alignment · initial conditions

Why Context Is Not the Solution

More tokens increase entropy, not reliability.

entropy expansion · domain mismatch · hallucination · 0-context architecture

Latent Space Navigation

State-space dynamics, not semantics.

dynamical systems · attractor regions · manifold constraints · trajectory control

LLMs as Transfer Functions

Not intelligence. Operators between domains.

transfer functions · domain mapping · control loops · operator perspective

Thinking Budget Optimization

Allocating reasoning compute across execution phases.

reasoning budget · phase allocation · planning vs execution · adaptive compute

The Automation Primitive: Domain Intelligence (To2D)

Transfer of Two Domains. The primitive underneath everything.

domain intelligence · domain resolution · To2D · structural foundation