The Binding Gap
An investigation into why LLMs often preserve semantic ingredients while failing at the exact relational binding that makes the answer faithful.
Archive
Public essays mirrored with provenance intact
This archive is filtered from the public Substack feed and rendered from local markdown snapshots. Each page keeps its source links, research folder, and raw files attached.
Operating Agents: Preface
The framing note for the Operating Agents course, introducing the chapter structure, blueprints, companions, and builder-first doctrine.
Operating Agents Bonus: Workflow Optimization and Evaluation
The optimization and evaluation bonus chapter of Operating Agents, focused on workflow optimization, measurement, and bounded self-improvement.
Operating Agents V: Trust Boundaries and Agent Safety
The safety chapter of Operating Agents, covering trust boundaries, attack surfaces, and architectural containment.
Operating Agents IV: Multi-Agent Boundaries and Handoffs
The multi-agent chapter of Operating Agents, focused on handoffs, transport, and when a second agent is actually justified.
Operating Agents III: Reasoning Scaffolds and Planning
The reasoning chapter of Operating Agents, covering scaffolds, planning, decomposition, and failure diagnosis.
Operating Agents II: Memory Systems and Authority
The memory chapter of Operating Agents, focused on memory layers, authority, retention policy, and evaluation.
Operating Agents I: Action Systems and Tool Use
The action chapter of Operating Agents, covering tool use, action surfaces, and the boundary between describing work and performing it.
The Illusion of the Swarm
A research-backed argument that many multi-agent systems are temporary routing workarounds rather than the final architecture.
The Moltbook Phenomenon
An analysis of Meta's Moltbook acquisition and what it signals about product strategy, distribution, and AI-native social software.
The Autonomy Tax
A thesis on why enterprise agents fail on control surfaces, auditability, and operations long before they fail on raw intelligence.
The Transformer Attractor
A study of why modern model design keeps converging back toward transformer-like structures across domains and workloads.
When the LLM Programs Its Own Thinking
An exploration of shared-state Jupyter workflows where an LLM acts less like a chat agent and more like a programmable REPL collaborator.
The Orchestration Paradigm: Issue 4 - The Reality
The closing issue in the Orchestration Paradigm series, focused on what survives contact with production constraints and evaluation.
The Orchestration Paradigm: Issue 3 - The Behavior
The third issue in the Orchestration Paradigm series, centered on behavior, coordination patterns, and failure modes inside agent stacks.
The Orchestration Paradigm: Issue 2 - The Factory
The second issue in the Orchestration Paradigm series, tracing how agent systems become factories of routing, tooling, and coordination.
The Orchestration Paradigm Series
The series introduction and framing piece for a four-part investigation into AI orchestration patterns and their production reality.
The Hardware Friction Map
A practical map of how hardware economics shapes model design, deployment choices, and the viability of ambitious architecture ideas.
What Actually Works: The Hardware Compatibility Filter in Neural Architecture (2023–2025)
The longer quantitative essay and dataset-backed argument for why hardware compatibility predicts which architectures reach production.
Autonomous AI Agents: Core Foundations and Recent Breakthroughs
A builder-first survey of foundation agent architectures, what changed in the literature, and how to choose practical defaults.
Neural Architecture Design as a Compositional Language
The inaugural Rooted Layers essay on architecture design as a compositional system rather than a loose collection of model families.
Welcome
The opening note for Rooted Layers and the public research archive that sits behind the publication.
No posts match the current filter.