Markdown Tidy
An experiment in structured context.
A minimal intervention to maintain clean input for both humans and AI agents.
Observability
Making agent reasoning transparent.
TryBase
Systems where AI judges, and humans calibrate.
Building experimental AI tools to explore autonomous reasoning and human-in-the-loop systems.
Shipping raw, failing fast, and iterating in public.
Experiments
We build small tools to test our hypotheses on human–system interaction.
Most will fail, but the logs will remain.
Notification Hell
An experiment in attention control.
A prototype resisting systemic interruption and reclaiming human attention.
(Soon) Context Stream
An experiment in reasoning transparency.
Tracing how agents form and adjust decisions.
Vision
Freedom is sustained by order.
We design structures that enable calm, deliberate action.
TryBase explores judgmental systems under human calibration—how we can trust AI agents with real-world decisions without losing control of the outcome.
Consistency
Measuring and preventing intent drift.
Delegation
Designing protocols for human oversight.
Roadmap
Q1 2026
Context Stream (Prototype)
Q2 2026
Observation API
Later
Human-in-the-loop Framework