# INCOPREA RESEARCH — EXPEDITION III
## THE PLAUSIBLE IMPOSSIBLE

**300 hypotheses · 30 research-ready refinements · 3 deep investigations · 1 atlas**

---

## The method that makes this different

Expeditions I and II generated a hypothesis first and asked for its consequence second — I organised
by subject, II by reasoning route. Expedition III inverts the order:

> **Start with a surprising outcome that established science does not forbid. Then work backward:**
> what conditions make it possible → what mechanism → what measurable consequence → what experiment
> decides it.

Nothing is admitted to the 300 without a mechanism and a kill criterion. The batch authors culled
from larger candidate sets (~32 → 25 in every batch) and recorded what they rejected.

---

## What is here

| Volume | File | Scale |
|---|---|---|
| **A** — 300 hypotheses, families 1–12 | `01_VOLUME_A/VOLUME_A_COMPLETE.md` | 300 entries, 188k words |
| **B** — 30 refinements, ranked | `02_VOLUME_B/VOLUME_B_COMPLETE.md` | 52k words |
| | **Total** | **≈ 247,000 words** |
| **C** — 3 investigations | `03_VOLUME_C/VOLUME_C_COMPLETE.md` | 3 results + code |
| **D** — atlas | `04_VOLUME_D/ATLAS.md` | ~2.2k words |

---

## The twelve outcome families

| # | Family | The surprise it starts from |
|---|---|---|
| 1 | Delayed / threshold-time | an effect that does not exist until a clock has run |
| 2 | Inverse response | a response that grows when the drive weakens |
| 3 | Disorder-improved order | precision that rises with noise or disorder |
| 4 | Window phenomena | an effect present only between two limits |
| 5 | Precursors | a signal arriving before the main event, without exceeding c |
| 6 | Measurement paradoxes | a measurement that improves when the apparatus is worse |
| 7 | Defect-borne properties | a property that exists only because the material is faulty |
| 8 | Gain from loss | amplification produced by the lossy element itself |
| 9 | Scale inversion | the large dictating the small |
| 10 | Function without mechanism | a function achieved with the wrong component |
| 11 | Mathematical surprises | an exact result hidden by a standard approximation |
| 12 | Relational phenomena | an effect that exists only in a comparison |

---

## The three deep investigations

| | Hypothesis | Result | Status |
|---|---|---|---|
| **C1** | Two identical barriers transmit $T^2/4$ at antiresonance | Exact law is $T^2/(2-T)^2$; resonance transmission is exactly 1. Verified against a transfer-matrix solution to 7 figures | **Derivation + verification**; the original number is only a weak-barrier limit |
| **C2** | Adding a pipe lowers a network's throughput | No inversion in either formulation, for any exponent 1–3. The Braess paradox needs *selfish routing*, not a variational flow | **Refuted as stated**, with the reason and the rescue conditions identified |
| **C3** | Short pulses maximise energy above critical coupling | Closed form; optimum $\to\kappa_0$ as $\kappa_0T\to\infty$ and $\kappa_e^*T\to2.513$ as $\kappa_0T\to0$. Critical coupling loses 94% at $\kappa_0T=0.05$ | **Derivation + verification** |

**One refutation, one correction, one new conditional result.** None is a claim about a measured device.

---

## Escaping two archives and two expeditions

The Phase 0 analysis (`00_ESCAPE/ESCAPE_ANALYSIS_III.md`) documents something specific: reading
Volume XIII (62 entries) and Volume XXIV (189 entries) of the Incoprea archive in full reveals that
**every entry is one of three rotating sentence templates** —

1. "X follows the *rate* of Y rather than Z"
2. "X is *set by the ratio* of A to B"
3. "X *needs a threshold* in the ratio of A to B"

The topic changes; the mechanism never does. All three templates are **banned and audited** in
Expedition III: the build script reports **0 of 300** uses of any of them.

Expeditions I and II (600 hypotheses) supply the second exclusion map; saturated mechanism families
are listed by ID, and every Volume A entry carries a prior-research section naming its nearest
collision.

**Coverage limitation, stated honestly:** two volumes read in full, one truncated catalogue (208 of
450), and the archive front pages — roughly 5% of 4,200 documents by count, though the template is
invariant across all 251 entries read. No complete duplication check was performed.

---

## What is genuinely new here

1. **An outcome-first generator.** The hypotheses could not have been produced by surveying a field,
   because they begin from a consequence and derive their field afterwards.
2. **A pattern-level escape from a published archive**, with the archive's templates identified and
   mechanically excluded.
3. **A recorded cull.** Every batch reports how many candidates were considered and why ~7 were
   dropped — including cases where the batch author discarded their own earlier draft (H156's
   original thermal-conductivity claim, reversed mid-draft).
4. **Three honest outcomes**, one of which is a refutation of a hypothesis that looked plausible
   precisely because it imported a famous result from another discipline.

---

## Reproducing

```bash
cd 05_SUPPORTING && python3 build_volume_a.py     # rebuild Volume A + QA
cd ../02_VOLUME_B && python3 build_volume_b.py    # rebuild Volume B + ranking
cd ../03_VOLUME_C/C1_barrier_identity && python3 c1_barriers.py
cd ../C2_braess_networks && python3 c2_braess.py
cd ../C3_pulse_coupling  && python3 c3_pulse.py
```

All deterministic, numpy only.

---

## Limitations, stated plainly

- **Volume A is idea generation.** The entries are mechanisms, not evidence. `[OBS]`-style claims
  name observation classes; prior-art status is marked unverified wherever a literature search was
  not performed.
- **Volume B refuted or narrowed a large fraction of its own candidates.** Twelve of the thirty were
  materially wrong in at least one claim — including H031 (C2), H182 (the "gain above lossless" claim
  fails globally), H166 (isotope exponent is the same as BCS), H255 (the stated theorem is not
  Keller's), H258 (the correction attaches to the wrong density), H207 (amplitude overstated by ~8
  orders), H202 (thickness scaling wrong in both geometries).
- **The three results are model results.** C1 and C3 are proved inside stated models; C2 is a
  refutation of a claim, not a discovery. No empirical claim is made.
- **The atlas maps the collection, not the world.**