Inquiry with consequences
We explore how private AI systems, creative tools, and autonomous workflows reshape thinking, culture, and meaningful work. Our research sits beneath everything we build — 348 briefs and counting.
Published Research
Agents need to see code as structure, not text.
We built a pure-Go tree-sitter runtime with 206 grammars to make that possible.
Read Part 1: Perception →
Agent Planning & DecompositionHow autonomous systems break complex work into executable steps.
And why most planning fails silently.
Read Part 2: Planning →
Autonomous Agent ArchitectureThe six layers every serious coding agent needs.
Perception, planning, execution, memory, governance, and trust.
Read the Series Introduction →
Governance for Autonomous SystemsDeclarative rules engines that keep AI systems trustworthy.
Cost caps, routing policies, escalation thresholds.
Read: Building an Arbiter →
Local-First AI InfrastructureWhy sovereign AI matters, what it costs, and how to build systems that compound.
Instead of expire.
Read: Sovereign AI for Studios →
Continuous Research SystemsHow we built a research system that runs around the clock.
On local infrastructure.
Read: Continuous Research →
- Local-first AI architecture patterns
- Autonomous agent governance via Arbiter
- Structural code analysis with gotreesitter