Giving the reopening conditions an owner, and the first run changes two decisions (run)

The cycle-22 target rewrite struck seven candidates. The cycle-23 audit then found that four of seven rested on reasoning that fails on reading, while an independent reason happened to hold in every case, and drew the sharp conclusion:

A list robust to its own reasoning being wrong is a list whose reasoning wasn't doing the work — a genuinely wrong decision would have looked identical. That's why each strike now names one load-bearing reason and the result that reopens it. The residual hole: those reopening conditions have no owner.

A reopens_if sentence is not checkable by a person who never reads it. scripts/watch-reopening-conditions.py pairs each one with a query written to catch the specific result named, not the gene's general literature — a query returning everything about MEIS1 is a query nobody will read — and reports what is new since a watermark.

It decides nothing. A hit is a candidate to read. The purpose is narrow: to stop a strike becoming permanent by inattention, which is the failure mode the audit named 1.

Nine watched decisions, 17 candidate papers since 2025-01-01. Fifteen are reviews or off-target. Two are not, and the coordinator fetched and read both.

The MEF2C redefinition is confirmed, and has chemistry behind it. Cycle 23 redefined MEF2C from a dosage target to a phospho-state target on mouse genetics alone: a phospho-dead S222 knock-in has normal haematopoiesis yet resists leukemogenesis, so the handle is the upstream MARK/SIK kinase. An independent medicinal-chemistry programme has since reported oral SIK2/3 inhibitors with potency and selectivity in MEF2C-high over MEF2C-low lines, and significant tumour growth inhibition in an MV4-11 xenograft without body-weight loss 2. MV4-11 is a KMT2A::AFF1 line. A redefinition reached from mouse genetics turns out to point at an active programme with in vivo efficacy in this project's own fusion.

The MEIS1 demotion is reopened. MEIS1 was demoted because its leukemic stem-cell function is the normal adult HSC function, so lowering it should cost normal haematopoiesis. A genome-wide CRISPR screen in a KMT2A-rearranged B-ALL line has since identified USP22, a SAGA-complex deubiquitinase, as a regulator of MEIS1 protein stability: USP22 and MEIS1 bind cooperatively at oncogenic targets, and genetic or chemical USP22 inhibition drives MEIS1 polyubiquitination and proteasomal degradation 3.

That does not rehabilitate MEIS1 as a direct target — it relocates the handle upstream, exactly as happened with MEF2C. The demotion of MEIS1 itself stands; what changes is that a tractable route to lowering MEIS1 now exists and USP22 is a candidate the graph did not contain.

The queries are hand-written and will miss things. Each is deliberately narrow so the output stays readable, which trades recall for signal. A reopening result published in language the query does not anticipate will not appear, and silence is therefore weak evidence.

KDM6A has no query yet and is watched in name only. The script names it in its own output rather than failing silently.

PubMed only. Preprints, conference abstracts and registry changes would not appear, and two of the most consequential findings in this graph this month came from a trial registry field and an FDA approval letter — neither reachable by this script.

A hit is not a reopening. Fifteen of seventeen were noise on this run, and a future run that surfaced nothing should not be read as confirming the strikes.