Cognitive architectures that bridge the gap between deterministic logic and aleatoric creativity.
Necessary Precision.
Artistic Chaos.
For the era of limitation to give way to abundance, human-centric systems must be an active and willing catalyst for their next logical successor.
Infrastructure of the agent age, not just adapted for it from human-limited heritage—languages, systems, and architectures engineered for machine-speed execution, unburdened by the necessity of human legibility.
- Deterministic Core Logic
- Stochastic Surface Layers
- Agent-Native Architecture
Algorithmic Curation
Proprietary algorithms to filter noise from signal, ensuring that every interaction within our software feels intentional and significant.
Fluid Interfaces
Interfaces that adapt to the user's cognitive load. Minimalist when focus is needed, rich and detailed during exploration phases.
The Human Loop
Technology that amplifies humanity, rather than replacing it. Systems designed to require human intuition as the final validation step.
Loomtown
The first artifact. A transitionary tool for massive scale agent orchestration. Loomtown operationalizes the 'Thinking at Scale' research, pursuing the reality of 1,000+ concurrent agents being coordinated with formal verification and distributed locking.
* Undergoing OSS Release Prep
1,000+ Concurrent Agents
Git Worktree Isolation
Optimistic Merging
Self-Correcting Verification Loops
Selected Papers
Thinking at Scale
Software engineering's foundational principles—Brooks' Law, Conway's Law, DRY—encode assumptions about human cognition that dissolve when the workforce shifts to 1,000+ AI agents. This paper argues the bottleneck migrates from code generation to specification and verification.
The Human Tax
We estimate that 30–50% of production codebases exist solely to make code legible to human minds. Agent-native systems should abandon text-based source code for structured representations (ASTs, graphs) and redesign the operational stack around machine-speed loops.