The Work Should Define the Agent
A response to Eno Reyes' question about agent boundaries, and why I think the durable unit is the work container rather than the agent team.
soul-os
17 SignalsThe Work Item Is the Product
What Amp's paved-road thesis and Truss Labs' prepared-work-item thesis reveal about Soul CLI, and why the next useful layer is a real case packet for agent work.
The Source Cabinet
A note on wiring Hermes, wallabag, and Soul together without turning every reading workflow into another chat transcript or another vault.
The Work Ledger: Turning Agent Work Into Verified Transactions
Why I added tasks, runs, worktrees, reviews, and plan-bound execution to Soul CLI, and how they turn AI-assisted development into an auditable local system.
Control Plane vs Code Plane in Soul OS
Why Soul keeps registry coordination separate from git, and why remote continuity should mean projection plus promotion rather than distributed writes.
Cross-Project Learning: Making the AI Persona Evolve
How to separate 'what I learned about the code' from 'what I learned about working with this person' so the AI gets better across projects.
Zero-Config AI: Moving Project Orchestration into the Registry
Why I stopped putting AI config files in every repo and moved project orchestration into a centralized registry instead.
Portable AI Personas: Decoupling Identity from Project
Why I stopped treating AI system prompts as per-project config files and started treating them as portable operating systems.
Unix-for-AI: Why Your AI's Memory Shouldn't Live in Git
How separating AI session state from your code repository into a dedicated registry follows the Unix philosophy and makes everything easier to debug.
Beyond Artifacts: Why AI Needs Persistent State
The shift from generating one-off code snippets to building persistent, registry-backed systems that maintain state across sessions.
The Bitter Lesson Applied to Agent Tooling
Why giving an AI agent one tool that does ten things leads to drift. Atomic, single-purpose tools are more reliable.
Auditing Soul OS Against the LLM Agent Lifecycle
Running the Soul OS architecture through a 7-step agent lifecycle audit. Where it's strong, and where the gaps are.
Solving Hallucinations in Technical Writing with Triangular Feedback
How using multiple models to cross-check documentation prevents AI-generated hallucinations from becoming permanent.
Applying the Bitter Lesson to Agent Architecture
What Vincent Sitzmann's 'Bitter Lesson of CV' means for how we should build AI agent systems — less hand-crafted modules, more end-to-end learning from traces.
From Artifacts to Systems: Adding Persistent State to Soul OS
The architectural pivot where Soul OS stopped generating ephemeral artifacts and started managing persistent state on disk.
Beyond the Cursor: A Primer on Agentic Coding
Notes from my Columbia University presentation on the three generations of AI in software engineering.
The Insight Engine: From Sequential Pipelines to Supervisor Architecture
Why sequential AI pipelines fail and how the Knowledge Supervisor pattern solves pipeline drift.