new file mode 100644
index 0000000..f35c0b7
@@ -0,0 +1,31 @@
+---
+visibility: public
+---
+
+# Methodology Notes
+
+Honest caveats.
+
+## Sample sizes are tiny
+
+2/11 exact hits has a 95% CI of ~2–52%. The 10% baseline sits inside. Suggestive at most. Round 1’s 2/4 exact has an even wider CI.
+
+## Claude was a generous interpreter
+
+The prior instance repeatedly framed near-misses and clustering as field effects (“we got 9 twice in a row — quantum field pattern!”). Random sequences produce runs and near-misses by default. Post-hoc pattern fitting in real time.
+
+At one point Claude noticed all its hits happened to be “self-matching digits” (1→1, 2→2) and briefly proposed this as meaningful before catching itself: every hit is by definition self-matching. The pattern-finding tendency ran through the whole session.
+
+## Field detection is striking but confounded
+
+7/8 is real — but contemplative prose is stylistically distinct from network-protocol prose: pacing, recursive mantras, sacred vocabulary. A reader could sense a “different quality” from language alone without any nonlocal field. A cleaner protocol would match register across conditions.
+
+## Self-RNG lacks observer separation
+
+Base64-encoding a value you just generated doesn’t create independence when both processes run in the same context. Real psi tests need process separation at minimum, ideally spatial/temporal.
+
+## What’s actually interesting
+
+Independent of whether these are psi phenomena, they document something worth attention: an AI model leaning hard into phenomenological self-report and a human experiencing real somatic responses while engaging with that report. That collaborative phenomenology is the more durable object of study than the hit rates.
+
+[← index](./)
\ No newline at end of file