Introduction: Making Mind Uploading / WBE a Measurable Problem

Key point

“The inputs and outputs look similar” is not enough for the kind of generation and internal causation WBE would require. That is why benchmark design is central.

When you get stuck on “Isn't the connectome enough?”

Wiring diagrams matter, but they do not by themselves satisfy even the lower bound of WBE requirements. If you want a literature-based explanation of what is still missing when synaptic efficacy, activity-dependent transcription / chromatin state, post-transcriptional RNA-state, cargo-transport / cytoskeletal trafficking state, perisynaptic ECM / perineuronal nets, thermal-state, ionic milieu / chloride homeostasis, bioenergetic / mitochondrial state, delay and myelination, neuromodulation, glia, cell-type labels, and intrinsic excitability / homeostatic set points are not captured, see Wiki: Why wiring diagrams alone are not enough and Wiki: Homeostatic plasticity and maintenance state.

When a thermodynamic paper sounds like the whole answer

At the WBE entry point, it is too early to treat one thermodynamic label as one solved layer. Mind-Upload separates Landauer lower bounds, biological energy budgets, irreversibility of coarse-grained neural time series, and model-based entropy-flow estimates. These can strengthen hypothesis formation, but on this site they remain auxiliary evidence unless the paper also discloses its signal route, estimator family, null / surrogate control, quantity type, and cost isolation. The shortest follow-up is FAQ: how to read thermodynamic claims plus Wiki: thermodynamic grounding basics.

Intrinsic-excitability evidence now gets its own route card

At the entry point, it is too coarse to read intrinsic-excitability evidence as one class. Yiu et al. (2014) and Hadzibegovic et al. (2025) are relative-excitability / allocation routes. Grubb & Burrone (2010), Kuba et al. (2010), Jamann et al. (2021), Fréal et al. (2023), and Benoit et al. (2025) are AIS / Na⁺-channel-state routes. O'Leary et al. (2014) and Hengen et al. (2016) are firing-rate set-point / recovery-control routes. Tallman et al. (2025) is a human clinical single-unit allocation route in epilepsy-associated hippocampal recordings, while Huber et al. (2013), Kuhn et al. (2016), Khatri et al. (2025), Fehér et al. (2026), plus Zrenner et al. (2018) remain living-human perturbation-conditioned proxy routes. Those rows do not share the same physiological locus, human evidence class, time axis, or observability ceiling. On this site, intrinsic-excitability claims now require a route card that fixes claim family, physiological locus, direct observable, time axis / intervention window, human evidence class / proxy class, and abstention boundary. The longer operating rule is in Wiki: intrinsic excitability / homeostatic-set-point route card.

Transcription / chromatin evidence now gets its own route card

At the entry point, the important correction is that allocation eligibility, acute time-course mapping, persistent stabilization programs, and locus-specific causal editability are not one evidence class, and neither are chromatin accessibility, histone chemistry, DNA-methylation control, higher-order looping, and locus editing one molecular object. On this site, readers are now directed to the Wiki: transcription / chromatin route card so claim family, molecular object, species, region, time window, assay, animal-level independence, human observability ceiling, perturbation status, and abstention boundary are fixed before a claim is promoted.

Post-transcriptional RNA-state is not just transcript abundance

At the entry point, it is now too coarse to read post-transcriptional RNA evidence as one class. Wang et al. (2015) is a splice-isoform route whose downstream object is chromatin / transcriptional control, Dai et al. (2019) is a splice-dependent receptor-balance route, Shi et al. (2018) and Li et al. (2025) are distinct m6A translation-versus-degradation routes, Peterson et al. (2025) is an RNA-editing route for homeostatic scaling, and Joglekar et al. (2024) is an atlas ceiling rather than a living-human in vivo readout. On this site, post-transcriptional claims now require a route card that fixes claim family, RNA control axis, assay / direct observable, downstream object, time window, human observability ceiling, and abstention boundary. The longer operating rule is in Wiki: post-transcriptional RNA route card.

Phospho-signaling evidence now gets its own route card

At the entry point, it is too coarse to read phospho-signaling evidence as one class. Lee et al. (2003) and Tomita et al. (2005) are phosphosite-specific plasticity gates, Havekes et al. (2016) and Vierra et al. (2023) are compartmentalized second-messenger routing, Altas et al. (2024) is region-specific phosphorylation with synapse-type relocalization in mouse and human samples, Rodriguez et al. (2025) is a single-site phospho-mutant causal memory intervention, and Biswas et al. (2023) is a human ex vivo phosphoproteome atlas. On this site, phospho-signaling claims now require a route card that fixes claim family, assay / direct observable, spatial or compartment scope, time window, causal leverage, and abstention boundary. The longer operating rule is in Wiki: phospho-signaling route card.

Local proteostasis evidence now gets its own route card

At the entry point, it is too coarse to read proteostasis evidence as one class. Frey & Morris (1997) and Shires et al. (2012) are tag / capture evidence, Govindarajan et al. (2011) is branch-level integration of protein-synthesis-dependent LTP, Fonseca et al. (2006) and Parker et al. (2025) are about synthesis-degradation / proteasome-capacity balance with memory consequences, Pandey et al. (2021) and Chang et al. (2024) are autophagy-linked plasticity routes, and Lee et al. (2022) plus Thomas et al. (2025) are about turnover-resistant persistence or a candidate tag substrate. On this site, proteostasis claims now require a route card that fixes claim family, integrative unit, direct observable, turnover / time window, controller / perturbation route, and abstention boundary. The longer operating rule is in Wiki: local proteostasis / synaptic-tagging route card.

Cargo-transport evidence now gets its own route card

At the entry point, it is too coarse to read cargo-transport evidence as one class. Park et al. (2006) and Correia et al. (2008) are about postsynaptic AMPAR / recycling-endosome delivery during LTP, Maas et al. (2009), Uchida et al. (2014), and Wong et al. (2024) are about transport-path state and local vesicle confinement, Nakayama et al. (2017), Liau et al. (2023), and Espadas et al. (2024) are about dendritic / synaptic RNA-granule organization and spine-targeted RNA support, de Queiroz et al. (2025) is about a distinct axonal RNA localization route in a mature in vivo memory circuit, and Aiken & Holzbaur (2024) is about presynaptic cargo delivery patterned by axonal microtubule dynamics. On this site, cargo-transport claims now require a route card that fixes claim family, cargo object, compartment scope, transport phase / state variable, trigger / time window, and abstention boundary. The longer rule is in Wiki: cargo-transport / cytoskeletal trafficking route card.

ECM / PNN evidence now gets its own route card

At the entry point, it is now too coarse to read ECM / PNN evidence as one class. Pizzorusso et al. (2002) is a plasticity-window reopening route, Frischknecht et al. (2009) is a receptor-mobility and short-term plasticity route, Nguyen et al. (2020) is a microglial ECM-remodeling route for memory consolidation, Alexander et al. (2025) dissociates CA2- versus PV-cell PNN roles across hippocampal memory tasks, Mehak et al. (2025) is an age-linked CA2 rescue route, and Lehner et al. (2024) plus Banovac et al. (2025) remain human ex vivo histology rather than living-human in vivo readout. On this site, ECM / PNN claims now require a route card that fixes claim family, matrix object / cell population, direct observable, controller / perturbation route, functional target, human observability ceiling, and abstention boundary. The longer operating rule is in Wiki: ECM / PNN route card.

Sleep replay evidence now gets its own route card

At the entry point, it is too coarse to read sleep architecture evidence as one class. Ngo et al. (2013) is a phase-locked auditory intervention in healthy humans, Baxter et al. (2023) adds a sleep-integrity boundary because oscillation gains did not translate into better memory, Whitmore et al. (2022) shows that cue benefit depends on uninterrupted N3 sleep, Schreiner et al. (2021) is endogenous scalp-EEG decoding around aggregated SO-spindle complexes, Schreiner et al. (2023) adds a respiration-linked physiology gate, Geva-Sagiv et al. (2023) is an intracranial closed-loop synchrony intervention, Schreiner et al. (2024) is human ripple-linked iEEG evidence, Whitmore et al. (2024) shows that sleep-disruption effects depend on memory age, and Duan et al. (2025), Shin et al. (2025), plus Deng et al. (2025) show item-selective, difficulty-selective, and time-window-specific variability rather than uniform overnight strengthening. On this site, sleep replay claims now require a route card that fixes preparation, event definition, timing / control policy, sleep-integrity / disturbance burden, NREM substate / physiology gate, memory target / selection or age, and abstention boundary. The longer rule is in Wiki: sleep replay route card.

Thermal evidence now gets its own route card

At the entry point, it is too coarse to read thermal evidence as one class. Hardingham & Larkman (1998) is a local operating-point physiology route, Moser et al. (1993) is a field-potential confound route, Long & Fee (2008) is a sequence-timing perturbation route, Owen et al. (2019) is a device-heating artifact route, Rzechorzek et al. (2022), Rogala et al. (2024), and Tan et al. (2025) are human passive / task-linked macro thermometry routes, while Tan et al. (2024) and Inoue et al. (2025) add human perturbation-conditioned thermal routes through severe heat exposure or intraoperative focal cooling. None of those human rows yet yields local thermal-controller ground truth. On this site, thermal claims now require a route card that fixes claim family, direct thermal observable, driver / perturbation route, time window, functional target, human route class, and abstention boundary. The longer rule is in Wiki: thermal route card.

Shared extracellular / electrical-state evidence now gets its own route card

At the entry point, it is also too coarse to read shared extracellular / electrical-state evidence as one class. Galarreta & Hestrin (1999) is a gap-junction coupling-network route, Anastassiou et al. (2011) is an endogenous-field / ephaptic route, Graydon et al. (2014) is an extracellular-space geometry / dilution route, Kilb et al. (2006) and Lauderdale et al. (2015) are osmotic extracellular-space contraction / edema routes, Burman et al. (2023) is an inhibitory-driving-force-regime route, Yang et al. (2024) is an activity-dependent electrical-synapse-remodeling route, Xie et al. (2013) is a sleep-linked interstitial-space route, Voldsbekk et al. (2020) is a human wakefulness-related diffusion-MRI ECS proxy clue, Örzsik et al. (2023) is a human sleep-conditioned higher-order diffusion / glymphatic clue, and Feld et al. (2026) remains a human perturbation-conditioned clue rather than a direct local readout. On this site, shared extracellular / electrical-state claims now require a route card that fixes claim family, direct extracellular / electrical observable, spatial regime, perturbation / calibration route, human evidence class, and abstention boundary. The longer operating rule is in Wiki: electrical-state route card.

Bioenergetic evidence now gets its own route card

At the entry point, it is also too coarse to read bioenergetic evidence as one class. Rangaraju et al. (2014) and Underwood et al. (2023) are about presynaptic ATP-linked support / respiration for transmission and memory formation, Rangaraju et al. (2019), Divakaruni et al. (2018), and Bapat et al. (2024) are dendritic positioning / fission and local plasticity support, Hu et al. (2025) is synaptic ATP-synthase nano-organization, and Vishwanath et al. (2026) is mitochondrial Ca²⁺-efflux tuning for long-term memory across species. Human observability is already split further: Ren et al. (2015) is a ³¹P metabolite / pH balance route, Ren et al. (2017) is a model-conditioned ³¹P MT exchange-flux route, Guo et al. (2024) is a whole-brain ³¹P NAD-content mapping route, Kaiser et al. (2026) is a localized functional ³¹P NAD-dynamics route, Karkouri et al. (2026) is a deuterium metabolite-mapping / absolute-quantification route, and Li et al. (2025) is a dynamic deuterium kinetic-rate imaging route. Those rows do not share one quantity type or one model burden. On this site, bioenergetic claims now require a route card that fixes claim family, compartment, direct energetic observable, quantity type / model burden, function target, and abstention boundary. The longer rule is in Wiki: bioenergetic / mitochondrial route card.

Synaptic-density PET evidence now gets its own route card

At the entry point, it is now too coarse to read SV2A / synaptic-density PET as one class. Naganawa et al. (2021) is about tracer and quantification-route validation, Johansen et al. (2024) is a healthy-human atlas route, Snellman et al. (2024) is a disease / risk-contrast route, Shatalina et al. (2024) is a task / cognition association route, Smart et al. (2021) shows that brief functional activation is not a momentary SV2A readout, and Holmes et al. (2022) shows that rapid ketamine response need not imply measurable SV2A change at 24 h. On this site, synaptic-density PET claims now require a route card that fixes claim family, tracer / quantification route, comparison design, functional target, and abstention boundary. The longer rule is in Wiki: SV2A / synaptic-density PET route card.

Presynaptic release machinery is not the same as synaptic-density or synapse count

At the entry point, it is now also too coarse to treat synapse count, regional SV2A density, and current presynaptic release machinery as one bundle. The site now keeps release-site number, docked-vesicle architecture, active-zone nanostructure / priming-site assembly, and current release competence separate from regional synaptic-density proxy rows. The longer literature-backed critique is in Wiki: Why wiring diagrams alone are not enough, and the audit-side stop rule is in Verification: latent-state error budget.

Glial substrate-routing evidence now gets its own route card

At the entry point, it is too coarse to read glial fuel support as one class. Suzuki et al. (2011) is about lactate-shuttle support, Silva et al. (2022) is about glial ketone-body export during starvation, Pavlowsky et al. (2025) is about intensive-learning glia-to-neuron fatty-acid flux, and Greda et al. (2025) is about apoE / sortilin-dependent lipid delivery and neuronal fuel-choice gating when glucose is limited. On this site, glial substrate-routing claims now require a route card that fixes claim family, supplier cell / neuronal sink, fuel object / carrier, regime trigger, transport route, and abstention boundary. The longer rule is in Wiki: glial substrate-routing route card.

Astrocyte evidence now gets its own route card

At the entry point, it is also too coarse to read astrocyte evidence as one class. Cahill et al. (2024) is about minute-scale cortical astrocyte-network encoding, Williamson et al. (2025) is about hippocampal ensemble recall, Dewa et al. (2025) is about multiday stabilization, Bukalo et al. (2026) is about amygdala fear-state representation, Villemagne et al. (2022) is a first-in-human MAO-B SMBT-1 target-validation route, Villemagne et al. (2022) is an AD-spectrum disease-context route, Hiraoka et al. (2025) is a brain quantification route, Mesfin et al. (2026) is a whole-body biodistribution route, Matsuoka et al. (2026) is a simplified / arterial-free SL25.1188 AD quantification route, Best et al. (2026) is a severity- and smoking-conditioned SL25.1188 AUD route, and Tyacke et al. (2018) plus Livingston et al. (2022) show that a human I₂BS PET route is a different target class whose readout can vary with region and impairment stage. On this site, human astrocyte-related PET is therefore treated as a target-defined and route-role-/quantification-/regime-conditioned proxy class, and astrocyte claims now require a route card that fixes claim family, direct astrocyte observable, perturbation route, functional target, human target / tracer family / route role / quantity type / quantification regime, and abstention boundary. The longer rule is in Wiki: astrocyte route card.

Astrocyte and clearance evidence do not yet collapse into one human controller readout

One remaining shortcut was to cite local astrocyte causal papers and human clearance proxies in one paragraph and let them sound additive. The primary literature does not support that shortcut. The direct observable in Villemagne et al. (2022) is tracer-defined MAO-B-related signal under a named quantification route, the direct observable in Lim et al. (2025) is respiration-conditioned awake-state CSF net flow, the direct observable in Yoo et al. (2025) is exercise-conditioned contrast-derived putative glymphatic influx / meningeal-lymphatic flow, the direct observable in Hirschler et al. (2025) is CSF mobility, the direct observable in Dagum et al. (2026) is overnight biomarker change interpreted through a multicompartment model, and the direct object in Cahill et al. (2024), Williamson et al. (2025), Dewa et al. (2025), Bukalo et al. (2026), and Kim et al. (2025) is local rodent causal control. On this site, those rows therefore stay separate unless a same-subject bridge, direct calibrator relation, and residual-latent budget are disclosed.

The myelin row also needs a route card

At the entry point, it is too coarse to read myelin evidence as one class. On this site, learning-dependent oligodendrogenesis, node / internode / periaxonal timing control, plasticity-brake evidence, remyelination-to-function recovery, human tract-scale transmission-speed estimation, and human myelin-sensitive or tissue-health-sensitive MRI routes such as myelin-water, MT-family, bilayer-sensitive, qT1 remyelination-sensitive, and T1w/FLAIR ratios are treated as different inferential objects. Therefore, a human myelin map, tract-speed estimate, or T1w/FLAIR ratio is not promoted here to solved biological timing-state, and functional recovery after remyelination is not read as proof that healthy myelin-state was completely restored. The longer operating rule is in Wiki: myelin / oligodendrocyte route card.

Ionic / chloride evidence now gets its own route card

At the entry point, it is also too coarse to read ionic evidence as one class. Glykys et al. (2014) is about local chloride set point, Heubl et al. (2017) is about activity-dependent KCC2 regulation, Ding et al. (2016) and Forsberg et al. (2022) are about interstitial / CSF ion composition linked to sleep-wake state, Byvaltsev et al. (2023) is about perisynaptic K⁺ clearance by reverse-mode KCC2, and Alfonsa et al. (2025) is about sleep-wake-history-related E_GABAA shifts that alter LTP induction. The human row is also split: Qian et al. (2012) is tissue-sodium mapping, Fleysher et al. (2013) is SQ+TQF-derived ISMF / ISC / ISVF, Rodriguez et al. (2022) is a repeatable normalized sodium density-weighted route, Azilinon et al. (2023) shows that TSC and short-component fraction f can diverge in the same disease context, and Qian et al. (2025) separates mono-/bi-T₂ sodium signals. On this site, ionic claims now require a route card that fixes claim family, direct ionic observable, spatial regime, perturbation / controller route, human quantity type / compartment model, and abstention boundary. The longer rule is in Wiki: ionic / chloride route card.

Neurovascular / BBB / blood-CSF barrier evidence now gets its own route card

At the entry point, it is also too coarse to read vascular and brain-fluid-barrier evidence as one class. Bell et al. (2010) is about adult pericyte-deficiency hypoperfusion and BBB breakdown, Kisler et al. (2020) is about acute neurovascular uncoupling, Pandey et al. (2023) is about pericyte-to-neuron memory signaling, Swissa et al. (2024) is about activity-dependent BBB modulation, and Mai-Morente et al. (2025) is about a pericyte capillary-diameter controller. The human BBB lane is also internally role-split: Morgan et al. (2024) is a method-comparison route showing that even ASL-derived BBB water-exchange is method-dependent, Padrela et al. (2025) is a healthy-adult lifespan reference route whose apparent gray-matter age effect disappears after CBF / ATT correction, Padrela et al. (2026) is a disease-burden contrast route showing lower Tex in SCD / MCI and moderate WMH burden while amyloid-group differences do not survive age / sex adjustment, and Chung et al. (2025) is a human tracer-specific BBB PET transport route. Mouse work by Ohene et al. (2019) further shows that multi-TE ASL exchange time is sensitive to AQP4 loss at the blood-brain interface, so the human water-exchange lane is not route-free. A distinct human BCSFB lane also exists and is internally role-split: Zhao et al. (2020) is an early choroid-plexus perfusion route-setting study, Sun et al. (2024) is a healthy-aging perfusion extension, Petitclerc et al. (2021) is a blood-to-CSF water-transport route, Anderson et al. (2022) is a choroid-plexus water-cycling route, Wu et al. (2026) is a repeatability anchor for apparent BCSFB exchange, and Petitclerc et al. (2026) is a simultaneous boundary-separation route that estimates both K_bl→GM and K_bl→CSF in one ASL acquisition. A third human barrier-side lane also exists in paired fluids: Farinas et al. (2025) used paired CSF and plasma proteomics in 2,171 individuals to derive individualized CSF/plasma ratios for 2,304 proteins, which is a paired-fluid protein-balance route rather than a direct BBB or BCSFB permeability scalar; the paper itself notes that ratio shifts can also reflect synthesis or degradation in either compartment, so the route cannot be read as transporter identity or route-free leakiness. On this site, neurovascular claims now require a route card that fixes claim family, biological locus, direct observable, human proxy role / evidence role, human quantity type / transport regime, crossed boundary / carrier class, interface sensitivity / dominant transport interpretation, validation / repeatability ceiling, and abstention boundary. The longer rule is in Wiki: neurovascular / BBB route card.

Do not fold local human ultrastructure into the in vivo proxy ladder

The safe reading here is stricter than simply saying that human evidence is getting richer. Lu et al. (2023) showed that preservation route changes extracellular-space retention and native geometry, Shapson-Coe et al. (2024) remained a rapidly preserved local surgical fragment, MICrONS Consortium et al. (2025) remained a sequential same-brain pipeline rather than same-time capture, and Dorkenwald et al. (2024) still required large proofreading effort at petascale. Therefore, this site treats local human ultrastructure as a destructive-route class and the living-human rows below as a separate in vivo proxy ladder.

Clearance / immune support is not passive cleanup

At the entry point, it is also too coarse to read clearance evidence as one class. Louveau et al. (2015) and Ahn et al. (2019) established meningeal-lymphatic drainage routes, Kim et al. (2025) showed that the meningeal-lymphatics-microglia axis can regulate synaptic physiology, and the human lane is no longer transport-only. Biechele et al. (2023) showed why TSPO is not a universal human activation-state meter, Wijesinghe et al. (2025) validated TSPO PET as a microglial biomarker in PSP, Horti et al. (2022) plus Ogata et al. (2025) established first-in-human CSF1R PET routes, and Yan et al. (2025) quantified COX-2 in healthy human brain. Fultz et al. (2019) then measured macroscopic CSF oscillations during human NREM sleep, Kim, Huang, & Liu (2025) measured parenchyma-CSF water exchange with MT spin labeling, Eide & Ringstad (2021) showed that sleep deprivation impairs molecular clearance in humans, Eide et al. (2023) linked intrathecal gadobutrol retention and population-pharmacokinetic CSF-to-blood clearance variables to different plasma biomarker patterns, Hirschler et al. (2025) measured region-specific CSF-mobility drivers, Lim et al. (2025) reported respiration-conditioned CSF net flow in awake humans while explicitly cautioning that plane-specific 2D PC-MRI net flow does not by itself represent true whole-brain bulk circulation, Yoo et al. (2025) reported exercise-conditioned contrast influx and parasagittal meningeal-lymphatic flow changes, and Dagum et al. (2026) linked sleep-active physiology to overnight Aβ / tau clearance to plasma in healthy older adults. On this site, drainage anatomy, microglia-related synaptic control, TSPO disease-context / validation-bounded PET, CSF1R route-setting PET, COX-2 enzyme-defined PET, macroscopic CSF oscillation, parenchyma-CSF water exchange, respiration-conditioned net-flow MRI, exercise-conditioned contrast influx, intrathecal tracer retention / CSF-to-blood clearance, human CSF-mobility MRI, and model-based human biomarker efflux are treated as different inferential objects. Therefore, a human CSF-mobility map is not read as direct flux ground truth, a respiration-conditioned net-flow estimate is not read as route-free whole-brain circulation truth, an intervention-conditioned contrast route is not read as a route-free baseline, an intrathecal-tracer route is not read as natural-sleep whole-brain clearance truth, a plasma Aβ / tau clearance model is not read as local immune-controller identification, and none of the PET rows is read as one generic human microglia-state scalar. The longer operating rule is in Wiki: clearance / immune route card.

Human immune support is not only a clearance-transport lane

This site now keeps two human lanes apart. Fultz et al. (2019), Kim, Huang, & Liu (2025), Hirschler et al. (2025), Eide et al. (2023), and Dagum et al. (2026) are transport-side human routes. By contrast, Wijesinghe et al. (2025), Horti et al. (2022), Ogata et al. (2025), and Yan et al. (2025) are target-defined neuroimmune PET routes. Those rows do not share one direct observable, one target, one tracer family, one time window, or one safe claim ceiling. On this site, immune evidence is therefore not read as one generic microglia meter and is not folded back into clearance transport.

Spectroscopy-derived human rows now need a quantity-type split

The remaining weakness in this ladder was that spectroscopy-derived human routes could still sound like one continuous advance. The primary literature does not support that shortcut. Lucchetti et al. (2025) defined the human metabolic connectome as a within-subject similarity matrix built from pairwise correlations among five metabolites across gray-matter parcels in 51 healthy participants, with replication in an independent sample of 13. Ren et al. (2015) used ³¹P-MRS in 12 subjects to estimate resting ATP synthesis, phosphorus-metabolite concentrations, and intra-/extracellular pH. Ren et al. (2017) then used band-inversion / magnetization-transfer modeling to estimate PCr→γ-ATP and Pi→γ-ATP exchange fluxes in the resting human brain. Guo et al. (2024) mapped whole-brain intracellular NAD content at 7 T, and Kaiser et al. (2026) detected task-evoked NAD⁺ dynamics in a functionally localized occipital voxel from 25 healthy volunteers. Li et al. (2025) used dynamic DMRSI plus a kinetic model to map CMR_Glc, CMR_Lac, V_TCA, and T_max in five healthy participants and reported repeat measurements only at the same operating point. Karkouri et al. (2026) then produced absolute HDO / Glc / Glx / Lac maps and rate maps at 7 T in 12 healthy volunteers plus five glioblastoma patients, with only two healthy post-glucose scans in the abstracted protocol. Ahmadian et al. (2025) further showed that deuterium dose materially changes downstream metabolite visibility, and Bøgh et al. (2024) showed that 3 T DMI repeatability depends on the named acquisition and time-point regime, with the best whole-brain repeatability at 120 min in that protocol. Those are not one inferential object. On this site, the 1H-MRSI row is read as a parcel-level biochemical similarity scaffold, the ³¹P rows as separate metabolite / pH, exchange-flux, NAD-content mapping, and localized functional NAD-dynamics proxy classes, and the deuterium rows as quantity-defined metabolite or kinetic-rate imaging under an explicit operating point, not as one merged spectroscopy ladder.

The measurement model also has to stay visible. Bhogal et al. (2020) showed that in vivo MRSI remains sensitive to low SNR, partial-volume effects, and extracranial lipid artifacts, Wright et al. (2022) showed that using averaged instead of voxel-specific metabolite T₁ corrections can bias maps, Baboli et al. (2024) showed that absolute quantification changes when tissue-water / relaxation correction is individualized, and Guo et al. (2025) showed that high-resolution 1H-MRSI metabolite-distribution mapping itself depends on acquisition / processing choices that manage spectral ghosting, aliasing, and low SNR. Therefore, on this site, a spectroscopy-derived human claim is not read without naming whether the object is static similarity, high-resolution metabolite distribution, energetic balance, deuterium absolute metabolite mapping / quantification, or kinetic rate imaging, together with the metabolite set, parceling or voxel unit, resolution plus PSF / partial-volume correction, water / lipid handling, spectral QC thresholds, and any blood-input or kinetic-model burden. For deuterium rows, the card now also asks for the dose, the time-point window, the field strength / coil route, and whether repeatability was shown only within the same setup or across protocols. The longer operational rules are in Wiki: metabolic-connectome route card and Wiki: bioenergetic / mitochondrial route card.

The tractography row now needs a route card

The weak point on this page was not that it called tractography a macro pathway prior, but that the phrase could still sound too stable, as if every tractography-derived connectome were the same inferential object once a modern pipeline had been applied. The newer route-audit literature argues against that shortcut at three different stages. Reveley et al. (2015) showed that superficial white matter can hide long-range cortical connections from roughly half of the cortical surface, and Schilling et al. (2018) showed that tractography endpoints are biased toward gyral crowns across deterministic and probabilistic algorithms, multiple diffusion models, and even very high-resolution data. Sarwar et al. (2023) then showed that filtering helps much less in complex human-like architectures than in simple bundles, He et al. (2024) showed that filtering can materially change structural laterality estimates, and Gajwani et al. (2023) showed across 40 pipelines and 44 group-representative reconstructions that hub location is highly variable and that hub connectivity correlates with regional surface area in 69% of assessed pipelines. McMaster et al. (2025) then showed that voxel size changes the resulting connectome and recommended resampling to 1 mm isotropic for robust comparisons, Bramati et al. (2026) showed on the same 3 T scanner with uniform processing that the q-space sampling scheme itself can change tractography outputs, Manzano-Patrón et al. (2025) showed that uncertainty should be propagated rather than hidden, and Zhu et al. (2025) improved reconstruction by fusing MRI with microscopy. Therefore, this site now asks readers to treat the tractography row as a route card: read acquisition / harmonization, cortical endpoint assignment, parcellation / weighting, uncertainty, calibration, and abstention before reading the graph. The longer operational rule is in Wiki: tractography route card.

Human proxy class is not the same thing as operational maturity

The strongest correction needed here was not to delete these rows, but to stop readers from treating them as equally mature. Johansen et al. (2024) built an SV2A atlas from 33 healthy participants calibrated against postmortem autoradiography, which is a real atlas resource but still a cohort-level synaptic-density proxy. Lucchetti et al. (2025) used 51 healthy adolescents plus an independent replication sample of 13 to derive a metabolic similarity matrix from five ¹H-MRSI metabolites, which is a biochemical organization scaffold rather than direct flux imaging. Ren et al. (2015) estimated resting ATP-synthesis-related, phosphorus-metabolite, and pH quantities in 12 participants, Ren et al. (2017) added a 7 T MT exchange-flux route with an explicit 5-pool model, Guo et al. (2024) mapped whole-brain intracellular NAD at 2.3 cc nominal resolution over roughly 51 min, and Kaiser et al. (2026) added a task-evoked 31P fMRS NAD⁺ route in 25 healthy volunteers. Meanwhile, Li et al. (2025) quantified glucose-transport and metabolic-rate maps in only five healthy participants using 7 T, custom dual-frequency coils, and blood input functions, and Karkouri et al. (2026) added absolute deuterium metabolite and rate maps with a separate calibration burden. Those spectroscopy rows are already enough to show that quantity type and acquisition burden diverge inside the same broad modality family. Meanwhile, Baadsvik et al. (2024) demonstrated myelin-bilayer mapping in two healthy volunteers on high-performance hardware. Likewise, Hirschler et al. (2025) introduced a specialized 7 T CSF-mobility route whose whole-brain rest maps were shown in 20 healthy younger individuals, with driver analyses reported in 11 of 24 total healthy participants. Dagum et al. (2026) combined a randomized crossover trial with 39 participants, an investigational wearable, and a compartmental model to infer sleep-linked clearance. Therefore, on this site, the ladder now has to be read along two axes at once: what variable class the route constrains and how specialized or deployment-limited that route still is.

Human proxy class, route maturity, and calibrator role are three different questions

Even when a living-human route is real, that still does not say how deployable it is or which hidden-state family it safely calibrates. Johansen et al. (2024), Lucchetti et al. (2025), Ren et al. (2015), Li et al. (2025), Karkouri et al. (2026), Baadsvik et al. (2024), Villemagne et al. (2022), Villemagne et al. (2022), Hiraoka et al. (2025), Mesfin et al. (2026), Matsuoka et al. (2026), Tyacke et al. (2018), Livingston et al. (2022), Best et al. (2026), Hirschler et al. (2025), and Dagum et al. (2026) do not calibrate the same latent variable. On this site, entry-level readers should therefore ask three separate questions before promoting a human route: what class is it, how operationally mature is it, and what does it actually calibrate. The longer technical rule is in Wiki: human proxy calibrator-role matrix.

PET rows need a measurement model

Within this ladder, PET-based rows are not readable from the modality label alone. Naganawa et al. (2021) showed that human SV2A PET quantification depends on the tracer, arterial-versus-reference quantification route, compartment model, and named scan window. Smart et al. (2021) showed that [¹¹C]UCB-J binding measures remain unchanged during brief visual activation even when tracer influx rises with blood flow, so momentary neural activity is not the same object as synaptic-density PET. Hansen et al. (2022) collated group-average receptor maps from more than 1,200 healthy participants, and Nakuci & Bansal (2025) then used those kinds of receptor / transporter densities as a modeling scaffold for spontaneous activity, so atlas priors remain normative rather than same-subject current-state readout. Wong et al. (2013) showed selected D₂-receptor target engagement by an administered drug, and Schlosser et al. (2025) showed that even a higher ketamine dose can return an informative null occupancy result for SERT rather than a release estimate. By contrast, Koepp et al. (1998), Erritzoe et al. (2020), and Miederer et al. (2025) used task- or challenge-linked binding changes as bounded release-sensitive proxies for selected dopamine or serotonin systems. Therefore, on this site, PET evidence is not read without its tracer, occupancy-versus-displacement design, quantification model, spatial scope, and time window.

Occupancy and displacement PET answer different questions

Wong et al. (2013) estimated whether an administered drug occupied a selected receptor target, and Schlosser et al. (2025) showed that the same occupancy framework can also return a route-specific null result rather than endogenous release. By contrast, Koepp et al. (1998), Erritzoe et al. (2020), and Miederer et al. (2025) used task- or drug-challenge-linked binding changes as endogenous release proxies within bounded windows. Those are not the same inferential object. On this site, occupancy PET is read as target engagement, while displacement PET is read as a challenge-limited release proxy. Neither is promoted to direct whole-brain neuromodulatory state.

Do not compress neuromodulation into one line

Mixed arousal proxies, local transmitter sensors, receptor / transporter atlas priors, occupancy PET, and displacement / release-sensitive PET are not one measurement class. On this site, they reduce different error terms and support different claim ceilings. The short rule is simple: do not promote a receptor map, an occupancy result, or a challenge-limited PET effect to the claim that the current whole-brain neuromodulatory state was identified. For the detailed ladder, see Wiki: neuromodulatory proxy ladder.

Neuromodulatory evidence now gets its own route card

At the entry point, it is too coarse to read neuromodulatory evidence as one class. Carro-Domínguez et al. (2025) is a mixed arousal proxy in human sleep, Neyhart et al. (2024) is a local transmitter-sensor route with explicit axon-activity and clearance constraints, Hansen et al. (2022) plus Nakuci & Bansal (2025) define a regional receptor / transporter atlas prior and modeling scaffold, Wong et al. (2013) plus Schlosser et al. (2025) are occupancy PET target-engagement routes, and Koepp et al. (1998), Erritzoe et al. (2020), plus Miederer et al. (2025) are challenge-limited displacement / release proxies. On this site, neuromodulatory claims now require a route card that fixes claim family, transmitter axis, direct observable, challenge or administered-drug route, time window / model burden, and abstention boundary. The longer operating rule is in Wiki: neuromodulatory route card.

Do not read multiple human proxy rows as automatic state completeness

A second correction follows from the same literature: even when several human proxy classes appear in the same argument, they still do not compose automatically into a state-complete readout. Shapson-Coe et al. (2024) is a local structural scaffold, Johansen et al. (2024) is a cohort-level synaptic-density proxy, Lucchetti et al. (2025) is a parcel-level biochemical scaffold, Ren et al. (2015) is a 31P metabolite / pH balance route, Ren et al. (2017) is a 31P MT exchange-flux route, Guo et al. (2024) is a 31P NAD-content mapping route, Kaiser et al. (2026) is a localized functional 31P NAD-dynamics route, Karkouri et al. (2026) is a specialized deuterium metabolite-mapping / absolute-quantification route, Li et al. (2025) is a 7 T dynamic deuterium kinetic-rate route, Baadsvik et al. (2024) is a specialized myelin route, and Hirschler et al. (2025) plus Dagum et al. (2026) are support-state routes with explicit model burden. These rows differ in quantity type, spatial unit, time window, quantification model, and operating burden, and the cited papers do not show that they can already be acquired in the same person, same session, same perturbation regime, with externally validated fusion. Therefore, on this site, proxy-rich human evidence is read as stronger than a single proxy row but still weaker than same-subject, cross-stack, externally calibrated state identification. The longer technical rule is summarized in Verification: Human Proxy Composition Card, Wiki: human-proxy composition rule, and Wiki: Human Proxy Composition and Route Maturity. That ranking is an inference from the measurement properties summarized above.

2026-03-21 rule: proxy bundles need an explicit composition audit

The correction that still remained after the March 2026 ladder update was operational, not conceptual. Readers could now see that Johansen et al. (2024), Lucchetti et al. (2025), Ren et al. (2015), Ren et al. (2017), Guo et al. (2024), Kaiser et al. (2026), Li et al. (2025), Karkouri et al. (2026), Baadsvik et al. (2024), Hirschler et al. (2025), and Dagum et al. (2026) constrain different state families, but the page still did not say when several rows may legitimately be promoted together. The primary literature supports a stricter rule: one row is a cohort-level synaptic-density atlas, one is a five-metabolite similarity graph in 51 adolescents plus 13-site replication, one is a 31P metabolite / pH balance route in 12 resting participants, one is a 7 T MT exchange-flux route, one is a whole-brain 7 T NAD-content map, one is a task-evoked 31P fMRS NAD⁺ route in 25 participants, one is a 7 T dynamic kinetic map in five participants, one is a 7 T deuterium metabolite-mapping / absolute-quantification route with its own calibration burden, one is a two-volunteer myelin proof-of-principle, one is a 20-person CSF-mobility rest-mapping route whose driver analysis was reported in 11 of 24 total healthy participants, and one is a model-heavy glymphatic clearance study in a randomized crossover trial of 39 participants. Those are not interchangeable pieces of one already-validated whole-brain state meter. The same multimodal literature also shows why agreement across rows is not enough by itself: Vafaii et al. (2024) found both common and divergent structure across simultaneous Ca²⁺ and BOLD, Chen et al. (2025) found tightly coupled global progression plus two distinct network patterns in simultaneous EEG-PET-MRI, Bolt et al. (2025) showed that a major global fMRI mode is substantially coupled to autonomic physiology, and Epp et al. (2025) showed that significant task BOLD changes can coexist with opposite oxygen-metabolism changes across many cortical voxels. The robustness literature makes the stop line narrower still: Finnema et al. (2018) showed route-specific SV2A PET test-retest constraints, Bøgh et al. (2024) showed route-local repeatability for 3 T deuterium metabolic imaging at its own operating point, Morgan et al. (2024) showed that even BBB water-exchange estimates can change materially across DP-ASL and ME-ASL within one cohort, Holiga et al. (2018) showed that common fMRI measures vary from poor to excellent in repeatability, Wirsich et al. (2021) showed that some simultaneous EEG-fMRI relations do reproduce across centres, Amiri et al. (2023) showed that full EEG+fMRI availability itself can be restricted to a subset, and Manasova et al. (2026) showed higher inter-modality disagreement in some clinically important groups even while multimodal performance rose overall. The key operational correction is now explicit: the same named quantity is not yet the same validated row. A repeatable 3 T DMI route does not automatically become interchangeable with specialized 7 T deuterium kinetic or absolute-quantification routes, and a BBB water-exchange estimate is still route-dependent when different ASL families give materially different numbers. On this site, a bundle of human proxy rows is promoted only after a Human Proxy Composition Card names the claimed latent variable, the stack that directly constrains it, the evidence relation across rows, the evidence role each row is actually allowed to play, the quantity type / model burden / acquisition burden, the cross-row nuisance audit, the external calibration route, the increment over the strongest single row, the residual latent families, the repeatability regime, the method-family non-equivalence, the cross-centre transfer window, and the row-overlap geometry / missingness policy. Operationally, those fields now compress to three promotion gates: robustness, common-driver / quantity-bridge separation, and increment over the strongest single row. Without that card, the bundle stays at proxy-rich but ceiling-limited human evidence. For the longer cross-stack critique, see Wiki: Human Proxy Composition and Route Maturity.

One more split is now required inside named proxy families. Naganawa et al. (2021) constrain an SV2A quantification route, Johansen et al. (2024) constrain a healthy atlas / baseline route, Snellman et al. (2024) constrain a disease / risk-contrast route, Shatalina et al. (2024) constrain a task / cognition association route, Smart et al. (2021) constrain an activation-null timescale boundary, and Holmes et al. (2022) constrain an intervention-response null at 24 h. Therefore, even at the entry point, SV2A PET is no longer treated as one reusable bundle row. The composition audit now has to name the family-internal comparison family before any shared synaptic-density role is inferred.

The same papers also do not license one shared evidence role. Johansen et al. (2024) is a healthy atlas / cohort-prior route, Snellman et al. (2024) is a cross-sectional risk-contrast route, Finnema et al. (2018) is a same-subject baseline / repeatability route, Smart et al. (2021) is a within-subject activation-change boundary, and Holmes et al. (2022) is a 24 h intervention-response boundary. On this site, the composition audit therefore also has to name whether each row is being used as a normative atlas / cohort prior, a cross-sectional contrast, a same-subject baseline readout, a within-subject change witness, or a perturbation-response witness. One evidence role is not allowed to stand in for another.

Same session is not yet the same state axis

Even a same-subject or same-session bundle can still mix different temporal objects. Lucchetti et al. (2025) is a static parcel-similarity scaffold, Guo et al. (2024) is a whole-brain NAD-content map, Kaiser et al. (2026) is a task-evoked localized NAD-dynamics route, Li et al. (2025) is a minutes-long kinetic mapping route, and Dagum et al. (2026) is an overnight perturbation-and-efflux route. Those rows do not all answer the same question at the same time axis. On this site, a proxy bundle therefore has to disclose effective time window / state-axis compatibility and physiological / perturbation regime compatibility before same-subject wording is allowed to raise the claim ceiling. The longer operational rule is in Wiki: Human Proxy Composition and Route Maturity and Verification: Human Proxy Composition Card.

Human tract-speed and myelin-sensitive MRI routes are not one reusable proxy row

The human timing-support lane has to be split before it can enter a proxy bundle. van Blooijs et al. (2023) constrain a tract-scale transmission-speed estimation route, which is a living-human timing-support advance but not a myelin-specific quantity. Arshad et al. (2017) compare MWF with calibrated T1w/T2w under repeat-scan reliability and concurrent-validity logic. Hagiwara et al. (2018) compare SyMRI, MT_sat, and T1w/T2w and show that the white-matter agreement pattern is not uniform across those methods. Baadsvik et al. (2024) then add a bilayer-sensitive route in only two healthy volunteers, while Galbusera et al. (2025) show that qT1, but not MWF or MTR, is sensitive to cortical remyelination in postmortem multiple-sclerosis cortex. Colaes et al. (2026) then show that T1w/FLAIR has only weak associations with MWF and supports a broader tissue-health reading rather than a myelin-specific one. Therefore, at the entry point, human timing-support evidence is not treated as one reusable row. A route has to be typed first as tract-scale transmission-speed estimation, MWF / calibrated T1w:T2w comparison, relaxometry / MT_sat comparison, bilayer-sensitive mapping, qT1 remyelination-sensitive pathology, or a T1w/FLAIR tissue-health-sensitive ratio before it is allowed to support a bundle claim.

What this table fixes at the entry point

The point of the comparison is not to downplay progress. The point is to stop three specific overreads before they spread upward through the site: quantity-type collapse (density = rate = similarity = mobility), deployment-maturity collapse (specialized proof-of-principle = routine same-subject observability), and fusion collapse (same-session agreement = one validated latent coordinate). The current primary literature supports progress on each row separately, but it does not support skipping those three audits.

Same-subject is still not same-state when the bridge is sequential

One more shortcut still has to be blocked at the entry point. A result can be same-subject or same-brain and still fail to be one state sample if the bridge runs from live measurement to later fixation / ex vivo follow-up or from one day / regime to another. Bosch et al. (2022) and MICrONS Consortium et al. (2025) show that strong correlative same-brain workflows still carry landmarks, targeted subvolumes, or local structure-function correspondences rather than one global state object. Gallego et al. (2020), Noda et al. (2025), Van De Ville et al. (2021), and Di et al. (2021) then show that latent manifolds, representational maps, and fingerprint features can remain stable on different timescales while raw units or amplitudes drift. Karpowicz et al. (2025), Wilson et al. (2025), and Wairagkar et al. (2025) further show that stable use across time can depend on alignment, recalibration, or only a short fixed-decoder horizon. Therefore, on this site, specimen identity is not allowed to stand in for state continuity. If a claim depends on treating a multi-stage bridge as one latent-state sample, it now also needs a Verification: State-Continuity Bridge Card that names the carried object / witness, the tolerance / failure rule, and the rescue route, in addition to the Human Proxy Composition Card or Destructive-Structure Route Card. The longer bridge-specific background is in Wiki: State-Continuity Bridge.

Proxy-rich evidence is still not the same as unique internal-state recovery

The remaining weak point in this ladder was that several strong rows placed side by side could still sound like a near-complete solution. The primary literature does not support that shortcut. Villaverde (2019) distinguished observability from structural identifiability, Prinz et al. (2004) showed that similar circuit activity can arise from disparate parameters, Rasero et al. (2024) showed that similar human activation patterns can still hide different macroscopic network states, and Beiran & Litwin-Kumar (2025) showed that even connectome-constrained networks remain degenerate until additional activity observations are added. Liu et al. (2025) then showed that practical identifiability depends on how data are collected, not only on how models are fit. Therefore, on this site, the human observability ladder answers which variable class becomes more constrained; it does not answer by itself whether the compatible internal-state family has collapsed to one explanation. That second question is handled separately in Verification: Identifiability Card.

The strongest causal papers and the best human routes are not the same ladder

A further correction is needed at the entry point. The papers that most strongly show that a maintenance-state family matters causally and the papers that most strongly improve living-human observability are often not the same papers, not the same species, and not the same spatial unit. Yiu et al. (2014), Hadzibegovic et al. (2025), Benoit et al. (2025), Hengen et al. (2016), Terceros et al. (2026), Dewa et al. (2025), and Bukalo et al. (2026) strengthen causal relevance in local rodent circuits. The strongest current human routes remain heterogeneous and route-limited instead: perturbation-conditioned excitability proxies such as Zrenner et al. (2018), target-defined astrocyte-related PET routes such as Villemagne et al. (2022), Tyacke et al. (2018), and Hiraoka et al. (2025), plus bounded clearance-side routes such as Hirschler et al. (2025) and Dagum et al. (2026). Those human rows already differ in target class, direct observable, time window, and model burden. Therefore, the primary literature does not license the shortcut from "causal importance was shown somewhere" plus "a human proxy improved somewhere" to "the responsible controller is now measured in humans". The composition rule that now enforces that stop line is spelled out in Wiki: Human Proxy Composition.

Human barrier MRI, astrocyte PET, neuroimmune PET, and clearance MRI still do not reveal the same controller object

The strongest human rows under this section still point at different biological objects. On the barrier side, Petitclerc et al. (2026) separated BBB-versus-BCSFB water transport with explicit compartment modeling in six healthy volunteers, while Chung et al. (2025) quantified tracer-specific BBB permeability and explicitly left human ground truth and test-retest for future work. On the astrocyte side, Villemagne et al. (2022) showed more than 85% selegiline blockade of ¹⁸F-SMBT-1 across the healthy human brain, which strengthens a target-defined MAO-B validation route rather than a memory-content route, and Hiraoka et al. (2025) then showed in six healthy participants with arterial blood sampling that scan window, reference region, and kinetic-model choice still materially shape how [¹⁸F]SMBT-1 is quantified. Tyacke et al. (2018) established a first-in-human I₂BS PET route with a different pharmacological competition profile, so even astrocyte-related PET already splits by target class. On the neuroimmune side, Biechele et al. (2023), Ogata et al. (2025), and Yan et al. (2025) keep TSPO, CSF1R, and COX-2 apart as different target-defined human rows rather than one generic inflammation scalar. On the clearance side, Hirschler et al. (2025) used a CSF-specific 7 T MRI mobility route in 20 healthy younger individuals and showed region-specific driving forces plus a separate CAA comparison cohort, whereas Dagum et al. (2026) used an investigational wearable plus a multicompartment model in a randomized crossover study with 39 analyzed participants to infer overnight brain-to-plasma biomarker efflux. These human rows therefore do not share one direct observable, one time axis, one quantification regime, or one controller identity. On this site, they can narrow barrier transport, target-defined astrocyte burden, target-defined neuroimmune burden, or transport-side support-state uncertainty, but they still do not identify the astrocyte ensemble, the local immune effector, or the synapse-specific maintenance logic highlighted by the rodent causal papers.

Even simultaneous multimodal human acquisition still needs a Fusion Card

Same-session acquisition narrows one kind of mismatch, but it still does not prove that the modalities share one externally validated biological state variable. Kothe et al. (2025) showed that synchronization middleware solves stream alignment rather than device-side delay truth, Wei et al. (2020) showed that EEG-fMRI fusion remains model-conditioned, and Vafaii et al. (2024) plus Chen et al. (2025) showed that simultaneous multimodal recordings can reveal both common and divergent structure across modalities. Bolt et al. (2025) and Özbay et al. (2019) further showed that low-frequency/global fMRI-linked fluctuations can also carry autonomic physiology. Therefore, on this site, even strong same-session multimodal human evidence is still read through a Fusion Card rather than as automatic state completeness.

A shared multimodal factor is not yet one solved state variable

This site now makes one more separation explicit at the entry page. A shared cross-modal component is a statistical object, not yet a biologically unique one. If a paper reports a coupled low-frequency mode or one common latent factor, the public claim still has to say whether that factor is being read as a shared neural candidate, a modality-specific residual left in the background, or a physiology-linked global factor. That disclosure now lives inside the Fusion Card rather than being left implicit under the word multimodal.

Nanoscale structure still needs a destructive-route audit

At the entry point, it is also too strong to read nanoscale, petascale, or same-brain EM as if the route had already preserved the native live state by default. Lu et al. (2023) showed that fixation route materially changes extracellular-space preservation and that even high-pressure freezing remains thickness-limited, Shapson-Coe et al. (2024) showed that a human nanoscale reconstruction is still a rapidly preserved local surgical fragment, and MICrONS Consortium et al. (2025) showed that same-brain function plus EM is a sequential local pipeline rather than simultaneous whole-state capture. Therefore, on this site, destructive ultrastructure is read through a preservation / registration / proofreading audit, not from resolution language alone. The longer rule is in Wiki: preservation / registration / throughput wall for destructive ultrastructure and Verification: Destructive-Structure Route Card.

Entry rule from this ladder

Do not collapse "human evidence exists" into one sentence. Structural scaffold, synaptic-density PET, receptor / transporter atlas priors, occupancy PET, displacement / release-sensitive PET, perturbation-conditioned sleep-homeostasis / plasticity proxies, state-gated perturbation routes, biochemical similarity scaffolds, ³¹P metabolite / pH balance, ³¹P MT exchange-flux, ³¹P NAD-content mapping, localized functional ³¹P NAD-dynamics, deuterium metabolite-mapping / absolute-quantification, deuterium kinetic-rate imaging, ionic proxies, thermometry, myelin maps, and clearance proxies are not interchangeable. On this site, each row reduces a different latent-state error term, and several of the higher-rung human routes still remain specialized, model-heavy, or small-cohort. None of them by itself upgrades WBE to state-complete or field-ready measurement. That last sentence is an inference from the measurement classes summarized above.

Do not promote perturbation-conditioned human routes to direct controller readout

Within this ladder, Huber et al. (2013), Kuhn et al. (2016), Fehér et al. (2026), and Zrenner et al. (2018) show that wake / sleep / nap history and ongoing EEG-defined state can change plasticity efficacy in humans. That is a real advance, but it still does not directly reveal AIS geometry, channel distribution, or the cell-specific recovery controller that produced the change. On this site, these rows are therefore read as perturbation-conditioned proxies, not as direct excitability ground truth.

Do not confuse hidden state with the measurement stack

The hidden states listed here are a list of important variables, not a list of variables currently observed. For what EEG, fMRI, spatial transcriptomics, Patch-seq, volume EM, and same-brain connectomics directly observe, and where each one reaches its claim ceiling, see Wiki: Measurement-stack observability and claim ceilings. The purpose is to stop the word “multimodal” from being read as if it already meant state-complete. In the same way, healthy-human MRS thermometry (Rzechorzek et al., 2022), task-linked thermal mapping (Rogala et al., 2024), and frontal-lobe thermometry (Tan et al., 2025) are important passive or task-linked macro thermal routes, while Tan et al. (2024) and Inoue et al. (2025) add bounded human perturbation-conditioned thermal routes through severe heat exposure and intraoperative focal cooling. None of those human rows, however, directly observes cell-specific temperature microgradients, synapse-level heating burden, or local thermal-controller state. Human five-metabolite 1H-MRSI similarity scaffolds (Lucchetti et al., 2025), high-resolution 1H-MRSI metabolite-distribution mapping (Guo et al., 2025), ³¹P metabolite / pH balance (Ren et al., 2015), ³¹P MT exchange-flux (Ren et al., 2017), ³¹P NAD-content mapping (Guo et al., 2024), localized functional ³¹P NAD-dynamics (Kaiser et al., 2026), deuterium metabolite-mapping / absolute quantification (Karkouri et al., 2026), and dynamic deuterium kinetic-rate imaging under blood-input and explicit kinetic modeling (Li et al., 2025) are important but distinct macro energetic routes. They still do not directly observe branch-local ATP reserve or mitochondrial positioning, and they do not collapse similarity, high-resolution metabolite distribution, metabolite balance, exchange flux, NAD content, local task dynamics, deuterium absolute metabolite mapping / quantification, and kinetic-rate imaging into one solved energetic object. Current human in vivo routes likewise do not directly observe isoform choice, m6A-reader engagement, or RNA-editing ratios across the whole living brain; specialized long-read atlas work such as Joglekar et al. (2024) is informative but still not a whole-brain in vivo human route. Current human in vivo routes likewise do not directly observe phosphosite occupancy, kinase/phosphatase balance, or compartment-specific second-messenger nanodomains across the whole living brain; ex vivo phosphoproteome atlas work such as Biswas et al. (2023) is informative but still not a comparable in vivo whole-brain human route. Human sodium MRI is also not one fixed ionic object: Qian et al. (2012) mapped tissue sodium, Fleysher et al. (2013) estimated ISMF / ISC / ISVF with combined SQ+TQF imaging, Rodriguez et al. (2022) reported repeatable normalized sodium density-weighted quantification, and Qian et al. (2025) separated mono- and bi-T₂ sodium signals. Those quantity types still do not directly observe cell-specific chloride concentration, KCC2 / NKCC1 balance, or local inhibitory reversal potential, and they do not collapse routine whole-brain intra- versus extracellular sodium partition into a solved problem. Current human in vivo routes likewise do not directly observe branch- or bouton-specific cargo pausing, motor engagement, or microtubule traffic state. Likewise, receptor / transporter atlas priors, occupancy PET, and displacement / release-sensitive PET constrain selected neuromodulatory systems, but they still do not directly observe the instantaneous whole-brain transmitter field.

How this hidden-state list becomes an operational rule

The March 2026 update added the Observability Budget to Verification in order to convert hidden-state criticism into an operational rule. For several living-human proxy rows used together, the site now also asks for a Human Proxy Composition Card so direct observables, evidence roles, same-subject relation, model burden, and residual latent states are not hidden behind the word multimodal or the phrase proxy-rich. For multimodal or atlas-prior results, the site additionally stacks a Fusion Card so that acquisition relation, synchronization, co-registration, fusion model, shared-vs-specific component disclosure, and abstention boundary are made explicit rather than implied by the word multimodal. For destructive ultrastructure claims, the site additionally asks for a Destructive-Structure Route Card so preservation route, registration scope, throughput burden, and omitted live-state families are not hidden behind the words petascale or nanoscale. From L1 upward, results are expected to report these boundaries together.

What this means in the introduction

The direct takeaway is that connectome-complete means progress on the structural scaffold, not that one may already declare emulation-complete. That is an inference from the literature above, which leaves maintenance-state variables as a separate problem. The detailed justification is concentrated in Wiki: Why wiring diagrams alone are not enough and Wiki: Homeostatic plasticity and maintenance state.

Three questions to ask when reading news or conference claims

1. Where is this on the L0-L5 ladder? Distinguish reproducible analysis, classification, intervention modeling, and identity claims first.
2. Is it only output matching, or does it survive condition changes? This prevents decode from being misread as emulate.
3. What would count as failure? Strong claims without a falsification condition should be read conservatively.

When you get stuck at the entrance to L4

The hard part of identity is not only the philosophical proper nouns. It is also deciding what should count as a continuity test. For the entry point to memory, values, learning, branching, and longitudinal continuity, see Wiki: Identity assessment and continuity tests.

When you get stuck on updates and multiple copies

Versioning, branch identifiers, and stop conditions become important when a system updates gradually through learning or when multiple branches are run from a common origin. For that distinction from the ground up, see Wiki: Update, branching, and stop rules.

When you get stuck on L3 closed loop

Even real-time success remains weak if delay, jitter, end-to-end return, and safe-stop logic are ambiguous. For that entry point, see Wiki: Closed loop, latency, jitter, and safe stops.

Low latency is not the whole L3 story

A fast loop can still be body-incomplete. Musall et al. (2019) and Stringer et al. (2019) showed that ongoing behavior shapes a large fraction of brainwide activity, Saleem et al. (2013) and Ravassard et al. (2013) showed that locomotion, optic flow, vestibular, and other sensory cues reshape cortical and hippocampal codes, Zelano et al. (2016) showed that nasal respiration modulates human limbic activity and memory, and Flesher et al. (2021) showed that tactile feedback improves a local bidirectional BCI loop. But the fast routes are not the whole boundary. de Quervain et al. (1998) and Oei et al. (2007) showed that glucocorticoid state can impair retrieval and reduce hippocampal / prefrontal retrieval activity, McCauley et al. (2020), Barone et al. (2023), and Birnie et al. (2023) showed that circadian timing and corticosteroid rhythm alter hippocampal plasticity, and Benedict et al. (2004), Reger et al. (2008), and Sherman et al. (2015) showed that insulin delivery or circadian-rhythm consistency can shift human memory or hippocampal activity. Therefore, Mind-Upload reads a closed-loop result safely only after the paper says which body / environment channels were preserved, substituted, or omitted, and which slow internal-milieu routes remained matched or latent.

The common substitution error

If a result is summarized only as “brain to text worked,” what may actually exist is a translator optimized to produce plausible text. To support a claim closer to WBE, the system must be checked under branching conditions: if the condition changes, does the internal behavior still match in the relevant way?

When you get stuck at the entrance to causal verification

Held-out accuracy, intervention, counterfactual, and perturbation-based verification have different strengths. For a plain-language route into that distinction, see Wiki: Counterfactual, intervention, and perturbation verification.

Important constraint

Whatever the deeper metaphysical truth may be, Mind-Upload uses measurable functional indicators as the common language. The philosophical consequence is postponed until enough empirical structure exists to constrain it.

How to read this section

Non-invasive decoding, invasive closed loop, connectomics, and large-scale simulation are not simply competing camps. They are complementary routes that close different gaps. The question here is not which route “wins,” but what level of claim each route can currently support.

2026-03-25 addendum: non-invasive language decoding is not one evidence class

The front-door comparison was still too coarse when it summarized all non-invasive language work through Tang (2023) alone. The current primary literature is narrower and more structured than that. Tang et al. (2023) is a within-subject fMRI semantic-reconstruction route with an explicit subject-cooperation requirement. Défossez et al. (2023) is a non-invasive M/EEG speech-segment retrieval route over 3 s windows from large candidate sets whose predictions primarily reflect lexical and contextual representations. d'Ascoli et al. (2025) is a known-word-onset word-decoding route that still depends strongly on protocol, with MEG > EEG and reading > listening. Ye et al. (2025) is a prompt-conditioned generative route where brain input improves over a permuted-brain control but does not erase prompt / LLM dependence. On this site, those are therefore treated as different evidence classes rather than one monotonic path to unrestricted thought reading.

2026-03-28 addendum: invasive language BCIs are not one operational route either

The invasive side was still too coarse if it only contrasted `speech BCI` against `whole-brain emulation`. The current primary literature supports a sharper split. Willett et al. (2023) strengthened same-session communication throughput and also exposed a bounded fixed-decoder slice by showing reasonable offline performance even without new-day retraining. Littlejohn et al. (2025) strengthened streaming throughput / expressivity. Wairagkar et al. (2025) strengthened instantaneous voice synthesis with silence fallback, but did not erase the distinction between fast voice output and long-horizon fixed decoders. Singh et al. (2025) instead strengthened a cross-subject transfer / initialization route, and Karpowicz et al. (2025) plus Wilson et al. (2025) strengthened adaptive stabilization / unsupervised rescue. On this site, those are no longer read as one monotonic invasive-BCI ladder.

2026-04-01 addendum: therapeutic invasive decoders are not one universal route

One more invasive shortcut had to be blocked at the entry page. Merk et al. (2025) combined invasive electrophysiology with whole-brain connectomic fingerprints to decode movement without patient-individual training across 56 implanted patients, and extended the same platform to emotion and seizure-related routes. That is a major advance in symptom-linked therapeutic decoding, but it is still not a subject-free universal decoder or a step that closes WBE-level state completeness. Oehrn et al. (2024) then showed that adaptive DBS can improve outcomes through patient-specific biomarker selection rather than one generic controller, and Zhu et al. (2026) showed that even eyes-open versus eyes-closed physiology can shift basal-ganglia oscillatory feedback signals enough that fixed thresholds risk confusing benign state change with pathology. On this site, a generalizable invasive decoder is therefore read as a connectomics-informed, symptom-/state-conditioned therapeutic-control route that still needs a named implant target, controller family, physiological-state guard, and programming burden, rather than as generic internal-state reading or emulation.

What this comparison really shows

The strongest support in the current primary literature is for better decoding, closed-loop improvement in local subsystems, structure-function mapping, and stimulus-conditioned digital-twin / connectome-constrained predictor families. The key question is therefore not route preference, but which evidence profile supports which rung of the claim ladder.

Do not collapse connectome progress into one rung

This site now separates at least five kinds of connectome progress: wiring atlas, same-brain local scaffold, human macro pathway prior / tractography connectome, connectome-constrained conditional predictor, and identifiability audit. Dorkenwald et al. (2024), MICrONS Consortium et al. (2025), tractography-validation papers such as Thomas et al. (2014) and Grisot et al. (2021), and Lappalainen et al. (2024) strengthen different parts of that ladder, while Beiran & Litwin-Kumar (2025) show why connectome alone still does not uniquely fix dynamics. On this site, the MICrONS-side `digital twin` language is therefore read as a stimulus-conditioned response model layered onto a sequential same-brain scaffold, not as direct current synaptic-state readout or one solved local twin. Therefore, “connectome progress” is not one claim level on this site.

2026-03-21 addendum: flagship connectome papers still solve different problems

The current connectome frontier is stronger than older summaries allowed, but its strongest papers still do not support one single inference. Dorkenwald et al. (2024) provide a whole-adult-fly wiring atlas, yet the authors still warn that false-negative and false-positive synapses remain and explicitly threshold connections for downstream analysis. Schlegel et al. (2024) then showed across three fly hemispheres that edges stronger than ten synapses or at least 0.9% of the target cell type's total inputs persist more than 90% of the time, which means weaker edges and fine cell-type claims still need a robustness audit rather than binary trust. MICrONS Consortium et al. (2025) provide a same-brain local structure-function scaffold in one mouse visual-cortex volume, but the EM route still followed in vivo imaging sequentially and remained a local visual-system pipeline. Gamlin et al. (2025) further show that even inside a large EM volume, transcriptomic Sst types were reached by morphology-based cross-modal prediction rather than by direct transcriptomic readout from the connectome itself. Lappalainen et al. (2024) showed that connectome structure plus task optimization can predict rich activity, but models with only cell-type connectivity predicted neural activity poorly. Finally, Beiran & Litwin-Kumar (2025) showed that recurrent networks sharing the same synaptic weights can still diverge strongly in dynamics when biophysical parameters differ unless additional recordings collapse the solution space. On this site, these are therefore read as different evidence classes, not as one monotonic countdown to state-complete WBE.

2026-04-04 addendum: petascale and same-brain connectomics now have a dedicated four-wall audit

The remaining weakness at the entry point was that preservation kinetics, sequential bridge burden, proofreading scope, and dynamical underdetermination were still spread across several pages. This site now concentrates that critique in Wiki: Why petascale connectomics still stops early. The short rule is that petascale, nanoscale, and same-brain are not one victory word here: they must still pass a preservation audit, a bridge audit, a completeness audit, and a dynamics audit separately.

Connectome-constrained predictors need a route card

Even a strong connectome-constrained predictor is still read here first as a conditional model / hypothesis engine. The safe reading depends on disclosing which structural prior was actually used, which parameters remained fitted, which task/state regime was tested, which mechanisms were omitted, and where the claim stops. Without that route card, this site does not promote a good activity-prediction result to unique internal-state recovery. The longer operational rule is in Wiki: conditional-model route card and Verification: Observability Budget.

The next bottleneck is not modality count alone

Recent primary literature makes one correction necessary even at the entry page. Villaverde (2019) separates observability from identifiability, Prinz et al. (2004) showed that similar circuit activity can arise from disparate parameters, Rasero et al. (2024) showed that similar human responses can still hide different macroscopic network states, Beiran & Litwin-Kumar (2025) showed that connectome-constrained networks remain degenerate until additional activity observations are added, and Liu et al. (2025) showed that practical identifiability depends on the data-collection policy itself. Therefore, after fixing the victory conditions, the next build step is to decide which measurement or perturbation would actually rule out the main alternative internal-state explanations.