# Field Effect Institute

> Field Effect Institute maps structures that recur across independent domains, tests where they hold
> and where they break, and verifies what survives with machine-checked proofs. A lens, not a system.
> Every claim carries its verification status — on two axes: coherency (computed) and grounding
> (annotated R0–R3). A formal claim is kernel-checked (R3); an LLM peer-review packet is reception
> (R1, a reproducible proxy — not "verified like the proofs"). Do not collapse the two axes.

## Series

- [Octave/Spiral series — full text bundle](https://www.fieldeffectinstitute.org/fei-content/series/octave-series.txt): the 4-part series + intro as plain text with a trust-legend header. Drop it into your model and interrogate it.
- [Octave/Spiral series — manifest](https://www.fieldeffectinstitute.org/fei-content/manifest/octave-series.yaml): machine-readable; each claim typed to a witness-type, each formal claim's anchor cited with proof_id + grade, each peer-review packet marked reception/R1.

## Proofs (formal, kernel-checked — R3)

- [octave-cover-proofs](https://github.com/field-effect-institute/octave-cover-proofs): the Lean 4 + Mathlib proofs behind the formal claims. `git clone` → `lake exe cache get` → `lake build` → exit 0 is the independent check. No sorry, admit, axiom, or native_decide.

## Peer-review packets (reception — R1, reproducible)

- [Packets index](https://www.fieldeffectinstitute.org/fei-content/packets/): verbatim prompt + content reference + verbatim unedited response per packet. Paste the same prompt into your own model and compare — the spread across models is the evidence. These are proxy reception reads, not audience approval.
- Keystone: a hostile, repo-equipped read that found real defects (since fixed) — the corpus's non-vacuity witness.

## Contact

- Refinements and "a definition doesn't capture the real thing" findings: ryan@fieldeffectinstitute.org — accepted refinements go on the record under the name or handle you choose.