Drawloom

Synaptic shuttle

Build AI systems
people can understand
and direct.

A contract-first harness for controlled, portable and inspectable agent work.

Explore the decisions View on GitHub
Clarity · constraint · composabilityOpen source for a more capable tomorrow

A visible line of reasoning

From questions to working systems.

Drawloom keeps the reasoning behind an agent system close to the code: principles shape investigations, decisions carry evidence, and implementations stay accountable to both.

Drawloom decision processA luminous network links principles, questions, investigations, decisions, evidence and implementation. Each label opens the corresponding public material.01PrinciplesWhat we care about02QuestionsWhat do we needto figure out?03InvestigationsHow do we exploreoptions?04DecisionsWhat will we doand why?05EvidenceWhat supportsthis choice?06ImplementationHow does itcome to life?
  1. 01PrinciplesWhat we care about
  2. 02QuestionsWhat do we need to figure out?
  3. 03InvestigationsHow do we explore options?
  4. 04DecisionsWhat will we do and why?
  5. 05EvidenceWhat supports this choice?
  6. 06ImplementationHow does it come to life?
Trace the architecture through public decisions and their supporting evidence.

The operating belief

Open by design.
Serious by default.

Powerful agent systems need more than capability. They need explicit boundaries, inspectable decisions and implementations that can move between hosts without hiding what matters.

Explore the open-source repository

From the journal

Why Drawloom?

Three attempts to give a doctor more time for medicine, and the lesson I nearly missed.

Read the essay