Research Program

Exogenesis

Exogenesis is a research program that investigates a specific failure mode in AI-assisted coding: the systematic skipping of intent manifestation before code generation. We test whether inserting a pre-execution intent discipline reduces drift, assumption leakage, and rework in agent output.

Thesis

The core argument: why execution outrunning intent is a structural flaw, not a user error.

Field Notes

Controlled experiments comparing baseline and Exogenesis-gated agent runs.

Drift Taxonomy

Nine recurring patterns of semantic drift observed in prompt-driven generation, plus legitimate divergence.

Vocabulary

Canonical terms for discussing disciplined AI development with precision.

Agent

An intent interrogation agent under development. Pre-execution reasoning, automated.

Start here

Read the Thesis → Field Note #01 → Drift Taxonomy → Vocabulary