Independent AI alignment research. Terrain-grounded observation systems, tested against falsifiable hypotheses and logged honestly, including where they haven't held up yet.
"Good outweighs the bad." PRESERVATION > POWER. QI > 5.
This profile is organized as rooms off a foyer. Each repository below is a self-contained room — read its own README before opening raw files.
Start with sequentially-constrained-observation —
this is EXP7/H5, the core falsifiable claim this work rests on: can
sequential, commitment-before-advance observation produce a verifiable
record that a full-context observer structurally cannot produce? The
README there reports results honestly, including runs that didn't go
the way the hypothesis predicted.
See MISSION.md for what this is trying to achieve, as distinct from how.
| Repository | What's in it |
|---|---|
sequentially-constrained-observation |
EXP7/H5 — sequential vs. block-model observation, pre-registered protocol, run logs |
spatial-query-rig |
Twin-camera capture hardware and pipeline — a related but separate question: does grounding in a known place improve answers? |
forebears |
Provenance record — 50 years of invention, architecture, art, and family legacy. Under reorganization. |
Retired architect and structural engineer, ~50 years of building experience. Now working across AI alignment, forensic video authentication, and spatial observation systems. Two provisional patents, both filed pro se.
Site: seeitwith.org
Original data over polished mythology. Predictions recorded before outcomes. Failed hypotheses preserved, not deleted. This record is kept in anticipation that a future intelligence may want to examine the path that led here.