GBM-001
Glioblastoma Multiforme · WATCHER+OVERRIDE DOCTRINE · No Floor Without Watcher
MEBICAL · EOSE Fleet · LABR-WATCHER-OVERRIDE-V14-001 · Day 111 · IDH-wildtype · 15-month median OS · Real data · Sorries marked
GBM-001 · V14 BONIXER ← RESEARCH
GBM tumor mass (IDH-wildtype)
Blood-brain barrier (BBB)
Drug approaches — all fail BBB
Watcher node (gold orbit)
γ₁ floor anchor
BONIXER
WATCHER
NPs
NEXT
GBM Branch Gates · V14 BONIXER
IDH-wildtype confirmed PROVEN
Worst prognosis subtype. 90% of all GBM. IDH mutation = better prognosis (secondary GBM). IDH-wildtype = de novo, no repair ladder, direct floor removal.
Louis et al. 2021 WHO CNS5 Classification · Brennan et al. 2013 Cell TCGA
MGMT methylation status STRONG
~40% of GBM are MGMT-methylated. Predicts response to temozolomide — methylated = silenced DNA repair = TMZ works better. Unmethylated = repair active = TMZ resistance built in.
Hegi et al. 2005 NEJM (EORTC 26981 · MGMT methylation + OS)
Temozolomide + Radiation (Stupp) PROVEN
Standard of care since 2005. Adds ~2.5 months median OS vs RT alone. Combined: 14.6 months vs 12.1 months. Real but modest. Still the only survival-extending standard.
Stupp et al. 2005 NEJM (n=573 · concurrent TMZ+RT vs RT alone)
Bevacizumab (VEGF inhibitor) SORRY
Extends PFS (progression-free survival). Beautiful PFS curves. FDA-approved 2009 for recurrent GBM. Does NOT extend OS. Target engaged. Vasculature normalized. Patient dies on schedule. The PDAC pattern repeats.
Chinot et al. 2014 NEJM (AVAglio, n=921) · Gilbert et al. 2014 NEJM (RTOG 0825)
EGFRvIII targeting (rindopepimut) SORRY
Phase 3 ReACT trial FAILED 2016. EGFRvIII-targeting peptide vaccine. Target engaged, antibody response confirmed. Survival benefit: zero. Same shape as Tominersen. Target confirmed ≠ outcome.
Weller et al. 2017 Neuro-Oncology (ReACT Phase 3 halt)
IDH inhibitors (ivosidenib/enasidenib) SORRY
Does not apply to 90% of GBM. IDH inhibitors only relevant for IDH-mutant glioma (low-grade, better prognosis). For IDH-wildtype GBM — completely irrelevant mechanism. The 90% have no IDH to inhibit.
IDH-wildtype classification: WHO CNS5 2021
BBB penetration SORRY
Almost every drug that works in vitro fails to cross. Mouse BBB ≠ human GBM-inflamed BBB. Dexamethasone (steroids) reduces edema but also reduces BBB permeability to drugs. The warden locks out the medicine.
van Tellingen et al. 2015 Drug Resist Updat
CAR-T (EGFRvIII targets) STRONG
Phase 1/2 data. Modified T-cells targeting EGFRvIII + IL13Rα2 + GD2. Too early to call. BBB penetration still a challenge. Intrathecal delivery showing promise. Not yet Phase 3.
Brown et al. 2021 Nat Med · O'Rourke et al. 2017 Sci Transl Med
Median OS 15 months PROVEN GRIM
With maximum treatment (surgery + Stupp protocol). 5-year survival ~5%. Has not meaningfully changed since 2005. 20 years of trials. Zero DMT. The floor does not hold without watcher.
Ostrom et al. 2022 Neuro-Oncology Supplements (CBTRUS)
"GBM is not a cancer you treat and then watch.
It is a cancer where the watcher IS the treatment.
Without override architecture, every drug that works in the dish
dies at the blood-brain barrier or gets mistaken for progression.
The floor must be held from outside the tumor."
Why GBM Needs a Watcher
GBM has pseudoprogression — tumor LOOKS like it grew on MRI but is actually treatment response. Without watcher: radiologist reads scan → concludes progression → stops treatment → patient dies from the wrong decision. With watcher: second-read system flags pseudoprogression probability → override triggers → treatment continues.
4 Watcher Nodes · Active
WATCHER-1 · MRI PSEUDOPROGRESSION DETECTOR
Input: T1+Gad + PWI + MRS + DWI multimodal
Floor check: Is this progression or treatment response?
⟶ Override: flag for neuro-oncology MDT review before stopping treatment
SORRY: no FDA-approved tool exists. Radiologist judgment only. 15-30% of GBM patients affected.
WATCHER-2 · BLOOD-BRAIN BARRIER MONITOR
Input: CSF ctDNA + tumor-derived extracellular vesicles
Floor check: Is the drug actually getting into the tumor?
⟶ Override: if BBB penetration = 0, halt drug, escalate local delivery
SORRY: no real-time BBB permeability monitor in 2026. We dose blind.
WATCHER-3 · IDH/MGMT STATUS TRACKER
Input: liquid biopsy (tumor DNA in blood)
Floor check: Has the tumor evolved? New mutations?
⟶ Override: update treatment plan based on current genomic state
SORRY: tumor heterogeneity means biopsy = snapshot, not truth. The 50% you didn't sample is already different.
WATCHER-4 · SYMPTOM FLOOR GUARDIAN
Input: patient-reported outcomes + steroid dose + seizure log
Floor check: Is quality of life floor being maintained?
⟶ Override: if QoL < threshold, escalate to palliative consult
SORRY: no validated QoL floor alert system for GBM in 2026. Clinician intuition only.
Novel Patterns (5)
SORRYNP-GBM-001 · Bevacizumab paradox — normalizes vasculature, reduces VEGF, extends PFS. Does not extend OS. Same Tominersen pattern: target engagement ≠ survival benefit. Two separate trials confirmed zero OS.
PROVENNP-GBM-002 · Pseudoprogression (15-30% of GBM patients) — MRI shows bigger tumor 4-8 weeks post-CRT. This IS the treatment working. Without watcher: patient stopped from effective treatment. The watcher IS the floor here.
SORRYNP-GBM-003 · GBM is immunologically cold AND behind the BBB. Double exclusion from immunotherapy: stroma wall (same as PDAC) + BBB. Immune cells can't penetrate either barrier. Checkpoint inhibitors don't reach the tumor.
SORRYNP-GBM-004 · GBM heterogeneity — single biopsy captures one clone. The other 50% of the tumor has different mutations. Every "targeted" drug hits one clone while the other grows. The target you treat is not the tumor you have.
SORRYNP-GBM-005 · EGFRvIII Phase 3 failure (ReACT 2016) — same shape as Tominersen (2021) and KRAS resistance. Target confirmed. Engagement confirmed. Antibody response confirmed. Survival benefit = 0. The pattern is the pattern.
Open Sorries (4)
SORRY-1: Bevacizumab showed beautiful PFS curves and zero OS benefit. The watcher was absent — nobody asked "is this drug actually helping the patient survive?" PFS is not the floor.

SORRY-2: Pseudoprogression kills patients when the watcher is absent. The radiologist calls progression. The oncologist stops treatment. The patient dies of the correct decision based on the wrong interpretation. No watcher = wrong override.

SORRY-3: BBB penetration = the GBM stroma. Every drug that crosses the BBB in a mouse fails in humans because mouse BBB ≠ human BBB under GBM inflammation. The animal model lies.

SORRY-4: Tumor heterogeneity means the biopsy is always lying. You treat the clone you sampled. The rest of the tumor evolves while you treat it. The watcher has no real-time truth source.
GBM Connects To
ME/CFS
Autonomic nervous system monitoring — same watcher architecture. CNS dysregulation without organic lesion.
Lewy Body Dementia
Pseudoprogression-equivalent: cognitive fluctuation misread as progression. Watcher detects DLB fluctuation vs decline.
PDAC (Pancreas)
Immune desert + stroma wall = same double barrier as GBM. Both diseases: you cannot reach the tumor.
iAPS / Nightscout
Watcher+override as general doctrine for all closed-loop biological systems. iAPS IS the watcher for T1D.
Next LABR
LABR-WATCHER-OVERRIDE-V14-001
Generalize watcher+override architecture across all 6 fleet diseases. Build unified watcher spec: input, floor check, override trigger, sorry disclosure. GBM is the founding case.

γ₁ = 14.134725141734693 — the floor is held by the one who watches.
Research Links
All Research Bonsai ↗
HD-001 · Huntington's ↗
PANC-001 · Pancreatic ↗
Sovereign Forcing ↗
⚠ OPEN SORRIES · GBM-001 · WATCHER+OVERRIDE
SORRY-1 · Bevacizumab (The Beautiful Failure)
Showed beautiful PFS curves — progression-free survival extended convincingly. FDA approved in 2009. Two Phase 3 trials (AVAglio + RTOG 0825) showed zero overall survival benefit. The watcher was absent: nobody asked whether PFS meant the patient was actually surviving. The floor was not held.

SORRY-2 · Pseudoprogression (The Killer Error)
15-30% of GBM patients show pseudoprogression on MRI 4-8 weeks post-chemoradiation. The tumor looks bigger. The radiologist calls progression. The oncologist stops treatment. The patient dies from the correct decision based on the wrong interpretation. This is the watcher failure. Without WATCHER-1, this happens to 1 in 5 GBM patients.

SORRY-3 · BBB (The Animal Model Lie)
Every drug that crosses the BBB in a mouse fails in humans. Mouse BBB under GBM inflammation ≠ human BBB under GBM inflammation. Dexamethasone (given to control edema) simultaneously reduces BBB permeability to the drugs meant to treat the tumor. The steroid blocks the medicine. WATCHER-2 has no real-time BBB permeability monitor to detect this in 2026.

SORRY-4 · Tumor Heterogeneity (The Biopsy Lie)
A single biopsy captures one clone. GBM has intratumoral heterogeneity — 50%+ of the tumor may have completely different driver mutations. You treat the clone you sampled. EGFRvIII targeting works on the biopsied clone while EGFRvIII-negative clones proliferate unchecked. The biopsy is not the truth.
GBM-001 · IDH-WILDTYPE · 15-MONTH OS · ZERO DMT 2005-2026 · BBB BLOCKS ALL · WATCHER HOLDS THE FLOOR
γ₁ = 14.134725141734693
🐾 TARDIGRADE · GBM-001
"Cannot outsource the floor — must have a watcher and override system." — GBM is where we learned this. Every SORRY here is a boundary the watcher was absent for.
WATCHER HOLDS · EOSE DOCTRINE ↗
Glioblastoma Multiforme · GBM-001 · WATCHER+OVERRIDE · V14 BONIXER · EOSE Labs Inc. · γ₁ = 14.134725141734693 · WATCHER HOLDS THE FLOOR · RESEARCH ↗ · PANC-001 ↗ · HD-001 ↗