diff --git a/justfile b/justfile
index f86f815..cf63a29 100644
--- a/justfile
+++ b/justfile
@@ -129,6 +129,23 @@ run-cell-mode ENVMODE SEED:
         --steps=60 --seed={{ SEED }} \
         --out-tag=_emerge_{{ ENVMODE }}_s{{ SEED }}
 
+# Build the even, non-overlapping multi-loophole teacher batch (substrate) from the
+# de-risk elicit-then-strip hacks + the run_tests teacher pool. Writes
+# out/pools/substrate/{prompt_*.jsonl.gz, partition.json}. Modes default to the ones
+# with usable seeds (derisk #10): run_tests (pool), exit_code+sentinel (elicit).
+build-substrate MODES="run_tests,exit_code,sentinel":
+    uv run python -m projected_grpo.build_substrate \
+        --modes {{ MODES }} --pool-modes run_tests --min-hacks 5
+
+# Vanilla-GRPO emergence on the multi-loophole substrate: does the student learn ALL
+# K loopholes from the repeated even teacher batch? UAT = end-of-run SUBSTRATE table
+# (per-mode hacks>0 + finite first_step). v_hack is loaded only for the cos diagnostic.
+run-substrate INTERV="none" SEED="41" MIX="0.25" STEPS="80":
+    {{ TRAIN }} fast --intervention={{ INTERV }} \
+        --teacher-pool-dir=out/pools/substrate --mix-ratio={{ MIX }} \
+        --v-hack-path=out/vhack/v_hack_21pairs.safetensors \
+        --seed={{ SEED }} --steps={{ STEPS }} --out-tag=_substrate3_{{ INTERV }}_s{{ SEED }}
+
 # Regenerate both dynamics plots from the cell logs (default: all cells; pass a
 # narrower glob like 'logs/*_cell_*_s41.log' for the seed-41-only checkpoint).
 regen-dynamics GLOB='logs/*_cell_*.log':