Research record / 14 September 2026
The research frontier.
What the complete at-home campaign changed, what it checked, and what it still cannot answer.
Release frontier-20260914-v1: full ledger and test records
The whole campaign, not just the highlights
The computational supplement now supports downstream research before the missing cell-surface measurement exists. It asks what an intervention would do if each specified mechanism and action law were true. It does not compute the missing observation or select a proven mechanism.
The original 80-work-package ledger closes with 76 scoped deliverables and four explicitly unmet outcomes whose preparation is complete. Those are work-package counts, not discoveries. The recorded goal execution time was about 2 hours 18 minutes; it is not total project labor, cumulative language-model usage, molecular-dynamics runtime or evidence that every result was invented in that interval.
This page accounts for every work package, including reused results, unsuccessful recovery, preparation and maintenance. It publishes selected reproducible science, not the private workspace: correspondence, medical records, collaborator attachments and internal machine paths are deliberately excluded.
More normal-channel protein does not always mean more electrical output
WT means wild type, the reference channel without R104Q. In an explicit capacity-limited scenario, adding WT increases its surface abundance but lowers total modeled current because the assumed mutant contribution falls more. The high mutant-activity setting is deliberately hypothetical, not an R104Q estimate. This refutes a universal inference from increased WT surface to increased total current.
A useful combination is not necessarily a synergistic combination
A factorial comparison measures baseline, mutant reduction alone, added WT alone and both together. Interaction is the combined change minus the sum of the two separate changes. Every tested burden/delivery family contains positive and negative examples. The sampled interaction ranges from about -0.28148 to +1.21774 untreated-WT reference-current units. Both illustrated combinations improve on baseline, but one has a negative interaction.
Evidence: versioned work-package and verification recordcomplete interaction results
An extra condition can separate parameters, but only under its stated model
For the shared-input surface law, proportional doubling from W=M=1 to W=M=2 with the same reference and unchanged coefficients can identify capacity and an aggregate partner penalty. It still cannot split every molecular cause. Recovery becomes unstable as the ratio of the two outputs approaches one. Separately, 64 assumed noise/floor scenarios compared eight candidate mixing ratios. The best sampled ratio ranged from 0.5 to 10 in the 48 informative scenarios; 16 equal-burden scenarios were indistinguishable at every ratio, and 76 individual rows failed the declared signal floor.
Evidence: versioned work-package and verification recordnoise-aware ratio results
The assumptions behind each route are now visible
Mutant reduction has to be evaluated alongside unintended WT reduction and the amount of residual mutant activity. In 45 sampled parameter rows, the selectivity map crosses the break-even boundary twice: there can be separated benefit regions rather than one required-selectivity cutoff. The RNA model gives a fraction-only synthesis ceiling only while production, translation and decay are constrained. Mosaic and turnover bridges distinguish the fraction edited from expression and steady-state abundance from the rates creating it.
Evidence: versioned work-package and verification recordselectivity map
A real sequence-annotation error was resolved
The nominated MUC2 site maps through the exact reference-transcript alignment to NM_002457.5 p.Tyr4524His, not either of the two older annotations. A genomic exon-concatenation coordinate had been transferred into a full transcript with different offsets. Reproducing the old table preserved its error; alignment resolved it. Separately, all 6,480 existing prime-editing candidates across 30 spacers passed the stated primer-binding, copied-product and nick-geometry checks. An 88-family comparison preserves heuristic ties instead of inventing a unique best editor.
Evidence: versioned work-package and verification recordexact MUC2 alignmentfull prime-editing audit
The protein-detection candidate is better specified
Lys-C, an enzyme that cuts proteins next to lysine, gives matched 26-residue WT/R104Q candidate peptides in the checked isoforms. Their calculated neutral mass difference is 28.04253351 daltons. In 169,634 versioned human-proteome records, the full WT sequence has 14 matches, all labeled SCN5A; the introduced mutant sequence has none. An eight-residue WT prefix also matches SCN8A, another sodium-channel gene. The checked 12-residue prefix avoids that ambiguity in this search.
Evidence: versioned work-package and verification recordproteome source and full search scope
The structural lead points inside the channel, not between two channels
A public structure search returned 80 experimental entities. Detailed inspection of selected candidates found a resolved counterpart of R104 in the related sodium channel NaV1.7 near its own voltage-sensing region. In PDB 7XM9, the mapped R99 and R174 atoms form an intrachain proximity of about 2.991 angstroms. This is a lead for internal communication, not an established interface between channel copies. A related calcium-channel peptide bound to calmodulin was qualified as a control; another promising-sounding record contained no peptide.
Evidence: versioned work-package and verification recordstructural sources, atom pairs and controls
The last beat can hide a different recovery history
The new experiment omitted the sixth scheduled stimulus while continuing to integrate the same interval. At the farther delivery mask, the next six stimulated windows changed from 010101 to 101010: three successes in either case, but the opposite final-beat label. The nearer mask retained 111111 in both schedules. Equal-mean uniform delivery retained 101010. A final-beat-only assessment would therefore give a different answer even though the farther mask's six-beat count did not improve.
Evidence: versioned work-package and verification recordfull tissue findings and verification scope
What the model actually computes
W and M are WT and mutant synthesis, each divided by the same reference WT-only synthesis. Surface abundance means protein at the cell membrane; electrical activity is the contribution per surface unit. They are not interchangeable.
q = b + lambda B S_W = (1+k) W / (1+k H_W) * W / (W+qM) S_M = u (1+k) M / (1+k H_M) A_W = W / (W+gamma uM) I = A_W S_W + a S_M
The assumed coefficients are survival u between zero and one, competition b, susceptibility lambda, capacity k, mutant activity a and coupling gamma. The latter five are nonnegative. The four partner-burden alternatives set B to zero, u, 1-u or u(1-u). None is fitted here as R104Q biology.
The three delivery laws set (H_W, H_M) to (W, M), (W, W+M) or (W+M, W+M). The middle choice attenuates mutant delivery by total input but not WT delivery; it is not a shared-capacity model. The separate WT penalty is not asserted to obey a conservation law.
At W=0, WT surface and WT activity are defined as zero. At W=M=0, total output is zero. At W=1 and M=0, WT surface, activity and current equal one. Current here is a nonnegative, additive protocol abstraction, not a simulated voltage trace.
An action changes W to alpha_W W + W_added and M to alpha_M M, with optionally declared changes to u, lambda or gamma. Other parameters remain fixed. The same action is applied to a separate WT-only control. Calling a parameter change a treatment does not show that a real intervention achieves it.
Exact grid, proofs and identification limits
The dose grid uses u=0.1, 0.5, 0.9; k=0, 2; a=0, 0.2, 1; b=1; lambda=2; gamma=1, across four burden models and three delivery laws. The generator specifies the action levels. Equivalent rows are not independent observations.
Under fixed parameters, proportional supply produces terms proportional to t/(1+At), so modeled current cannot decrease. Isolated reductions of the stipulated harmful susceptibility or coupling also cannot decrease it. Those are properties of the defined actions, not universal treatment effects.
For the joint-input law, let s1 and s2 be WT surface outputs at W=M=1 and W=M=2 and R=s2/s1. Then k=(2-R)/(4(R-1)) and q=(1+k)/((1+2k)s1)-1 for the admissible domain. Recovery is unstable near R=1 and does not split q into its component mechanisms.
For productive RNA, R=P f/d and synthesis T=tR. If only productive fraction f changes, the fold ceiling is 1/f0 for a positive baseline. Changes in production P, translation t or decay d require additional bounds. Zero production or translation remains a zero-output exception.
Model implementationAction implementationExact dose generatorIndependent equation reconstructionFull methods appendixScientific source-note hashes
Every work package, accounted for
Open any entry for its original requirement, accepted scope and remaining dependency. “Scoped deliverable” includes a verified predecessor, a negative search or a prepared plan when that was the original requirement. It does not mean a new successful experiment.
01-6: Shared computational foundation
A versioned model separates synthesis, survival, surface abundance and electrical activity. Forecast-only records cannot satisfy measured-observation gates. Reference conversions preserve empty-partner versus WT-partner normalization. Numerical error, sensitivity, model disagreement and measured uncertainty have different labels.
Scoped deliverable verified01. Define a downstream input contract.
Original requirement: Specify WT and mutant synthesis, survival, surface abundance, activity, units, reference conditions and provenance. Deliverable: a schema distinguishing learned, assumed, simulated and measured quantities.
Accepted scope: Per-quantity conditional schema, not empirical ingestion.
Record: frontier-20260914-v1 / work package 1. Download the public ledger.
Scoped deliverable verified02. Add absolute-expression capacity.
Original requirement: Extend the ratio-based model to distinguish increasing WT, reducing mutant and increasing both. Deliverable: explicit equations with WT-only and zero-input limits; no invented fitted capacity.
Accepted scope: Original conditional mathematical/design deliverable only. No calibrated biological rates, empirical intervention effect, or mechanism identification. See CHECKPOINT_05_ACCEPTANCE.md.
Record: frontier-20260914-v1 / work package 2. Download the public ledger.
Scoped deliverable verified03. Add a surface-to-current component.
Original requirement: Keep WT and mutant activity separate and permit distinct gating hypotheses. Deliverable: current scenarios with a documented domain, not inferred current from abundance scores.
Accepted scope: Original conditional mathematical/design deliverable only. No calibrated biological rates, empirical intervention effect, or mechanism identification. See CHECKPOINT_05_ACCEPTANCE.md.
Record: frontier-20260914-v1 / work package 3. Download the public ledger.
Scoped deliverable verified04. Build downstream adapters.
Original requirement: Connect supplement outputs to forecast-only interfaces. Preserve empty-vector versus WT-partner normalization and never bypass a measurement gate. Deliverable: validated conversions and rejection tests.
Accepted scope: Authenticated frozen-source conversion and integrated forecast example; no measured quantities.
Record: frontier-20260914-v1 / work package 4. Download the public ledger.
Scoped deliverable verified05. Establish independent numerical tests.
Original requirement: Check units, limiting cases, conservation where stipulated, derivatives and finite changes against a second implementation. Deliverable: saved test results and counterexamples for any failed assertion.
Accepted scope: Original conditional mathematical/design deliverable only. No calibrated biological rates, empirical intervention effect, or mechanism identification. See CHECKPOINT_05_ACCEPTANCE.md.
Record: frontier-20260914-v1 / work package 5. Download the public ledger.
Scoped deliverable verified06. Separate uncertainty types.
Original requirement: Distinguish numerical error, parameter sensitivity, model disagreement and empirically calibrated uncertainty. Deliverable: an uncertainty/reporting contract that prohibits calling arbitrary grid ranges biological confidence intervals.
Accepted scope: Only original computational/design/review deliverable; no empirical validation, clinical claim, backup execution or external release.
Record: frontier-20260914-v1 / work package 6. Download the public ledger.
07-15: Intervention models
The executed action laws cover mutant reduction with WT leakage, added WT, proportional and preferential output, survival, partner burden and functional coupling. Every contrast retains a treated-WT control. The combination study compares all four constituent conditions instead of treating improvement as synergy.
Scoped deliverable verified07. Mutant-selective silencing, route 6.
Original requirement: Model dose-dependent mutant reduction and WT leakage. Deliverable: benefit/reversal boundaries across mechanisms, including partial versus complete reduction.
Accepted scope: Only original computational/design/review deliverable; no empirical validation, clinical claim, backup execution or external release.
Record: frontier-20260914-v1 / work package 7. Download the public ledger.
Scoped deliverable verified08. Added WT, route 7.
Original requirement: Hold mutant supply fixed and vary WT supply using item 02. Deliverable: absolute response curves with matched WT-background comparisons and explicit capacity dependence.
Accepted scope: Original conditional mathematical/design deliverable only. No calibrated biological rates, empirical intervention effect, or mechanism identification. See CHECKPOINT_05_ACCEPTANCE.md.
Record: frontier-20260914-v1 / work package 8. Download the public ledger.
Scoped deliverable verified09. Proportional upregulation, route 13.
Original requirement: Increase both alleles while retaining their ratio. Deliverable: supply-dependent surface/current responses; an unchanged ratio must not be mistaken for unchanged output.
Accepted scope: Original conditional mathematical/design deliverable only. No calibrated biological rates, empirical intervention effect, or mechanism identification. See CHECKPOINT_05_ACCEPTANCE.md.
Record: frontier-20260914-v1 / work package 9. Download the public ledger.
Scoped deliverable verified10. Preferential WT upregulation.
Original requirement: Sweep selectivity and total output separately. Deliverable: selectivity-capacity break-even maps, not a claim that a real intervention achieves those changes.
Accepted scope: Only original computational/design/review deliverable; no empirical validation, clinical claim, backup execution or external release.
Record: frontier-20260914-v1 / work package 10. Download the public ledger.
Scoped deliverable verified11. MOG1/trafficking modulation, route 10.
Original requirement: State how the intervention changes export, survival or burden before computing its effects. Deliverable: mixed-genotype response scenarios and treated-WT controls.
Accepted scope: Original conditional mathematical/design deliverable only. No calibrated biological rates, empirical intervention effect, or mechanism identification. See CHECKPOINT_05_ACCEPTANCE.md.
Record: frontier-20260914-v1 / work package 11. Download the public ledger.
Scoped deliverable verified12. Folding correction, route 4.
Original requirement: Test increased survival under competing partner mechanisms. Deliverable: conditions under which stabilization helps or worsens the modeled endpoint; do not assume a stability score establishes a folding lesion.
Accepted scope: Original conditional mathematical/design deliverable only. No calibrated biological rates, empirical intervention effect, or mechanism identification. See CHECKPOINT_05_ACCEPTANCE.md.
Record: frontier-20260914-v1 / work package 12. Download the public ledger.
Scoped deliverable verified13. Turnover modulation, historical candidate 14.
Original requirement: Separate synthesis, delivery and removal. Deliverable: residence-time sensitivity and prospective contrasts. Do not reinstate the superseded route as validated.
Accepted scope: Original conditional mathematical/design deliverable only. No calibrated biological rates, empirical intervention effect, or mechanism identification. See CHECKPOINT_05_ACCEPTANCE.md.
Record: frontier-20260914-v1 / work package 13. Download the public ledger.
Scoped deliverable verified14. Coupled-gating perturbation, route 12.
Original requirement: Combine surface scenarios with independently stated functional coupling. Deliverable: contrasts distinguishing surface loss from activity loss, including effects on WT controls.
Accepted scope: Original conditional mathematical/design deliverable only. No calibrated biological rates, empirical intervention effect, or mechanism identification. See CHECKPOINT_05_ACCEPTANCE.md.
Record: frontier-20260914-v1 / work package 14. Download the public ledger.
Scoped deliverable verified15. Combination interventions.
Original requirement: Only after single-intervention models pass, examine a small justified set such as silencing plus WT addition. Deliverable: interaction maps with explicit assumptions, not an exhaustive combinatorial sweep.
Accepted scope: Conditional computation only; independent review accepted. Item39 zero-cap domain qualification corrected and fourth test added/passed after review. No empirical efficacy or calibrated probabilities.
Record: frontier-20260914-v1 / work package 15. Download the public ledger.
16-21: Robustness and decision value
A 96-group matrix retains the signs and counterexamples from all 9,072 dose rows. Bounded mathematical proofs establish specific monotonic properties; explicit witnesses defeat broader benefit claims. A two-level, common-reference experiment identifies an aggregate penalty and capacity only within one specified model.
Scoped deliverable verified16. Across-model intervention comparison.
Original requirement: Identify conclusions shared by the stipulated models and conclusions that reverse. Deliverable: a robustness matrix, not a forced treatment ranking.
Accepted scope: Conditional computation only; independent review accepted. Item39 zero-cap domain qualification corrected and fourth test added/passed after review. No empirical efficacy or calibrated probabilities.
Record: frontier-20260914-v1 / work package 16. Download the public ledger.
Scoped deliverable verified17. Find explicit reversal witnesses.
Original requirement: For each purported robust benefit, seek admissible counterexamples. Deliverable: reproducible witness parameters or a bounded proof with its exact domain.
Accepted scope: Original conditional mathematical/design deliverable only. No calibrated biological rates, empirical intervention effect, or mechanism identification. See CHECKPOINT_05_ACCEPTANCE.md.
Record: frontier-20260914-v1 / work package 17. Download the public ledger.
Scoped deliverable verified18. Identify influential missing quantities.
Original requirement: Determine which assumptions control intervention ranking or sign. Deliverable: a prioritized information-needs map; no probability-weighted value-of-information claim without justified probabilities.
Accepted scope: Original conditional mathematical/design deliverable only. No calibrated biological rates, empirical intervention effect, or mechanism identification. See CHECKPOINT_05_ACCEPTANCE.md.
Record: frontier-20260914-v1 / work package 18. Download the public ledger.
Scoped deliverable verified19. Test intervention-specific identifiability.
Original requirement: Ask whether proposed abundance/current observations distinguish the new intervention models. Deliverable: identifiable combinations and unresolved alternatives beyond the already completed baseline counterexamples.
Accepted scope: Original conditional mathematical/design deliverable only. No calibrated biological rates, empirical intervention effect, or mechanism identification. See CHECKPOINT_05_ACCEPTANCE.md.
Record: frontier-20260914-v1 / work package 19. Download the public ledger.
Scoped deliverable verified20. Extend the activity-bound application.
Original requirement: Use correctly referenced surface and separately assumed current scenarios to forecast possible bounds. Deliverable: conditional bounds with denominator and background checks, never a measured activity estimate.
Accepted scope: Only original computational/design/review deliverable; no empirical validation, clinical claim, backup execution or external release.
Record: frontier-20260914-v1 / work package 20. Download the public ledger.
Scoped deliverable verified21. Define stopping rules.
Original requirement: Set when additional parameter sweeps, model variants or simulations cease to change a decision. Deliverable: a stop/continue checklist for each executed branch.
Accepted scope: Reviewed design/decision document only; no assay validation, calibrated power or actual lab execution.
Record: frontier-20260914-v1 / work package 21. Download the public ledger.
22-30: Fluorescence and experiment design
Experimental proposals were separated from documented agreement without publishing correspondence. Total fluorescence, intact protein, surface protein and current remain separate observables. Actual synthesis need not follow DNA mixing ratio. Controls cover reporter maturation, background, burden and shared-reference covariance. Noise-aware ratio and sample-size sensitivity are prepared, but not calibrated assay power.
Scoped deliverable verified22. Reconcile the proposed experiments with meeting evidence.
Original requirement: Separate documented agreement, remembered discussion and later computational proposals. Deliverable: an evidence-graded experiment list; do not call all proposals “laboratory's plan.”
Accepted scope: Evidence-graded reconciliation of existing correspondence audit and late-call record; not a fresh complete transcript or inbox review.
Record: frontier-20260914-v1 / work package 22. Download the public ledger.
Scoped deliverable verified23. Specify observable requirements.
Original requirement: Distinguish total GFP, total WT protein, WT surface, mutant surface and current. Deliverable: a readout-to-inference table, building on existing gates rather than duplicating them.
Accepted scope: Reviewed design/decision document only; no assay validation, calibrated power or actual lab execution.
Record: frontier-20260914-v1 / work package 23. Download the public ledger.
Scoped deliverable verified24. Resolve exact construct configuration.
Original requirement: Seek sequence-bearing maps, tag/linker placement, isoform and H558 background. Deliverable: a verified construct record or explicit unresolved fields; reference models remain usable as reference models.
Accepted scope: Explicit unresolved construct fields; no claim that actual laboratory stocks were authenticated.
Record: frontier-20260914-v1 / work package 24. Download the public ledger.
Scoped deliverable verified25. Model DNA-to-synthesis mismatch.
Original requirement: Test expression imbalance independently of nominal plasmid ratio. Deliverable: sensitivity of the predicted ordering to actual synthesis ratios.
Accepted scope: Reviewed design/decision document only; no assay validation, calibrated power or actual lab execution.
Record: frontier-20260914-v1 / work package 25. Download the public ledger.
Scoped deliverable verified26. Design nuisance controls.
Original requirement: Evaluate how reporter maturation, background and expression burden could mimic the new predictions. Deliverable: an experiment-specific control matrix, not another generic dimming counterexample.
Accepted scope: Reviewed design/decision document only; no assay validation, calibrated power or actual lab execution.
Record: frontier-20260914-v1 / work package 26. Download the public ledger.
Scoped deliverable verified27. Select discriminating ratios.
Original requirement: Compare candidate conditions by separation of model predictions under stated noise. Deliverable: a compact proposed panel and conditions where discrimination fails.
Accepted scope: Candidate comparison under declared noise, not calibrated power or an approved panel.
Record: frontier-20260914-v1 / work package 27. Download the public ledger.
Scoped deliverable verified28. Write the prospective decision ledger.
Original requirement: For each candidate observation, state what it supports, refutes or cannot distinguish. Deliverable: dated, versioned falsification criteria before outcome data are available.
Accepted scope: Reviewed design/decision document only; no assay validation, calibrated power or actual lab execution.
Record: frontier-20260914-v1 / work package 28. Download the public ledger.
Scoped deliverable verified29. Calibrate power and uncertainty.
Original requirement: Obtain preparation-level variance, covariance, backgrounds and allocation information. At home, make sensitivity curves; deliver a real sample-size calculation only when inputs justify it.
Accepted scope: Assumed preparation-level power sensitivity only; actual allocation remains pilot-dependent.
Record: frontier-20260914-v1 / work package 29. Download the public ledger.
Scoped deliverable verified30. Prepare phenotype-driven follow-up.
Original requirement: Map outcomes to trafficking, turnover, interaction or functional follow-up. Deliverable: a staged decision tree; actual assays remain lab work, not computationally completed measurements.
Accepted scope: Reviewed design/decision document only; no assay validation, calibrated power or actual lab execution.
Record: frontier-20260914-v1 / work package 30. Download the public ledger.
31-36: Editing and oligonucleotides
Existing base-, prime- and RNA-editing designs were audited rather than replaced by another design sweep. All 6,480 prime-editing candidates passed the stated local sequence/product/nick checks. A full transcript alignment corrected the MUC2 consequence. The selectivity and mosaic bridges expose why sequence discrimination and fraction edited are not functional rescue.
Scoped deliverable verified31. Base-editor design audit, route 1.
Original requirement: Check reference context, intended edit, editing-window assumptions and bystanders in existing candidates. Deliverable: a versioned candidate comparison with unsupported assumptions flagged.
Accepted scope: Versioned existing base-editor reference/window/bystander audit, not efficacy.
Record: frontier-20260914-v1 / work package 31. Download the public ledger.
Scoped deliverable verified32. Off-target analysis audit.
Original requirement: Reconcile existing specificity results, reference assemblies, paralogs and tool settings. Deliverable: reproducible computational specificity evidence, not proof of biological off-target absence.
Accepted scope: Bounded inherited specificity workflow audited and reproduced; no exhaustive all-build/all-alignment certification. MUC2 correction retained separately.
Record: frontier-20260914-v1 / work package 32. Download the public ledger.
Scoped deliverable verified33. Prime-editing design audit, route 3.
Original requirement: Verify existing pegRNA/nicking configurations and sequence provenance. Deliverable: a candidate panel with computational rationale and unresolved empirical efficiency.
Accepted scope: All 6480 existing candidates verified against local reference for sequence/product/nick configuration; 88-family panel and exact reproduction. No empirical efficiency or safety.
Record: frontier-20260914-v1 / work package 33. Download the public ledger.
Scoped deliverable verified34. RNA-editing design audit, route 2.
Original requirement: Review target accessibility assumptions, intended/bystander edits and transcript specificity. Deliverable: a source-grounded candidate table.
Accepted scope: Source-grounded RNA candidate/proxy specificity audit, not biochemical specificity or accessibility validation.
Record: frontier-20260914-v1 / work package 34. Download the public ledger.
Scoped deliverable verified35. Silencing-sequence selectivity bridge.
Original requirement: Connect computational discrimination assumptions to the WT-leakage model in item 07. Deliverable: required selectivity ranges; do not equate sequence prediction with measured knockdown.
Accepted scope: Only original computational/design/review deliverable; no empirical validation, clinical claim, backup execution or external release.
Record: frontier-20260914-v1 / work package 35. Download the public ledger.
Scoped deliverable verified36. Editing-to-functional-rescue bridge.
Original requirement: Distinguish fraction edited, allele output and function. Deliverable: mosaic correction scenarios rather than treating editing percentage as current recovery percentage.
Accepted scope: Only original computational/design/review deliverable; no empirical validation, clinical claim, backup execution or external release.
Record: frontier-20260914-v1 / work package 36. Download the public ledger.
37-42: RNA, splicing and source recovery
Native RNA, minigene and cDNA evidence were reconciled without multiplying incompatible scores. The exact public ParSE software deposit and publisher files were checked; missing per-barcode PSI inputs were not recovered. APC score reconstruction retains a precise source dependency. Productive-transcript ceilings now state the production, translation and decay assumptions they require.
Scoped deliverable verified37. Native-RNA versus cDNA evidence map.
Original requirement: Identify which observations address splicing, RNA abundance or protein behavior. Deliverable: a denominator-aware evidence table; no multiplying incompatible assay scores.
Accepted scope: Completed predecessor distinguishes native RNA, minigene and conditional cDNA contexts and incompatible denominators.
Record: frontier-20260914-v1 / work package 37. Download the public ledger.
Scoped deliverable verified38. Recover ParSE replicate mapping.
Original requirement: Look for source-group keys, original replicate files or processed per-barcode replicate tables. Deliverable: verified mapping or a bounded negative search; no bulk reprocessing with invented labels.
Accepted scope: Bounded negative search and authenticated current public-blob comparison, as explicitly permitted by original deliverable. Missing PSI inputs remain missing; no inferred replicate labels.
Record: frontier-20260914-v1 / work package 38. Download the public ledger.
Scoped deliverable verified39. Productive-transcript flux model, route 5.
Original requirement: Separate production, splice fractions and decay/NMD assumptions. Deliverable: conditional ceilings with explicit cell/context limitations.
Accepted scope: Conditional computation only; independent review accepted. Item39 zero-cap domain qualification corrected and fourth test added/passed after review. No empirical efficacy or calibrated probabilities.
Record: frontier-20260914-v1 / work package 39. Download the public ledger.
Scoped deliverable verified40. Recover APC normalization provenance.
Original requirement: Identify the transformation, controls and raw-to-processed mapping needed for historical score interpretation. Deliverable: reproducible transformation or a precise unresolved-source statement.
Accepted scope: Precise unresolved normalization-source statement, as explicitly permitted by original deliverable. No exact linear-current reconstruction claimed.
Record: frontier-20260914-v1 / work package 40. Download the public ledger.
Outcome unmet / preparation complete41. Recover missing source supplements/exports.
Original requirement: Target the specific O'Neill/cohort/calibration materials still unresolved in current audits; verify they are not already present before requesting them. Deliverable: real source files, provenance and endpoint qualifications.
Accepted scope: The intended outcome remains unmet. Useful preparation and alternatives are complete.
Still missing: Authenticated original ONeill supplemental workbook and specified cohort raw exports not recovered.
Reopen when: A genuinely new legitimate source location or author-supplied original becomes available.
Usable alternative: Use already qualified source-scale summaries and conditional scenarios; retain provisional transcription labels.
Record: frontier-20260914-v1 / work package 41. Download the public ledger.
Scoped deliverable verified42. Evaluate protected-data access feasibility.
Original requirement: Determine whether needed cardiac transcript data are accessible through authorized channels. Deliverable: access requirements or a public-data alternative, not bypassed access controls.
Accepted scope: Authorized-access feasibility and alternatives, not acquisition of protected reads.
Record: frontier-20260914-v1 / work package 42. Download the public ledger.
43-48: Mechanism prediction and structure
The claim matrix distinguishes destabilization, abundance, assembly, surface effects and electrical dominance. No qualified new independent endpoint labels were found in the bounded update. Resolved homolog contacts support an intrachain structural lead, not an established two-channel interface. Background, isoform and accessory-protein effects remain distinct from undocumented stock identity.
Scoped deliverable verified43. Audit mechanistic claim strength.
Original requirement: Separate predicted destabilization, abundance, assembly, surface effects and functional dominance. Deliverable: a claim-to-evidence matrix that prevents one endpoint becoming another.
Accepted scope: Scoped endpoint/claim matrix, not universal historical claim repair.
Record: frontier-20260914-v1 / work package 43. Download the public ledger.
Scoped deliverable verified44. Seek genuinely independent endpoint labels.
Original requirement: Identify new NTD abundance/partner-function labels not used for gating or model selection. Deliverable: versioned coverage and overlap audit; no new calibration on recycled examples.
Accepted scope: Versioned bounded independent-label coverage audit with negative search result; no new calibration.
Record: frontier-20260914-v1 / work package 44. Download the public ledger.
Scoped deliverable verified45. Recover the historical full nomination score table.
Original requirement: Check specific new source locations for Paper 9's missing artifact. Deliverable: authenticated historical output or a documented distinction from any fresh scoring run.
Accepted scope: Historical-output recovery distinction; missing full nomination table remains missing.
Record: frontier-20260914-v1 / work package 45. Download the public ledger.
Outcome unmet / preparation complete46. Extend predictor benchmarking only with a new testable question.
Original requirement: Require independent labels, versioned models and held-out evaluation. Deliverable: a benchmark that could reject the proposed transfer; retain the existing failed partner transfer.
Accepted scope: The intended outcome remains unmet. Useful preparation and alternatives are complete.
Still missing: No new independent-label held-out benchmark executed.
Reopen when: Independent endpoint-compatible labels and a new falsifiable transfer question are identified.
Usable alternative: Preserve the completed failed partner transfer and bounded negative coverage audit; no recycled-label calibration.
Record: frontier-20260914-v1 / work package 46. Download the public ledger.
Scoped deliverable verified47. Qualify NTD interaction geometry.
Original requirement: Search source evidence for relevant resolved contacts or testable structural hypotheses. Deliverable: geometry provenance and limitations; no passing off a predicted pose as established assembly.
Accepted scope: Bounded geometry qualification, including resolved homolog intrachain contacts and related-class control; no target intermolecular interface established.
Record: frontier-20260914-v1 / work package 47. Download the public ledger.
Scoped deliverable verified48. Background/isoform interaction analysis.
Original requirement: Keep Q1077, H558 and beta1 context separate where evidence permits. Deliverable: context-specific assumptions or comparisons, not an inferred actual backbone or effects from identical scenario labels.
Accepted scope: Conditional background/configuration register, not actual backbone or empirical interaction.
Record: frontier-20260914-v1 / work package 48. Download the public ledger.
49-54: Molecular simulation
Existing runs and systems received individual interpretation and admission dispositions. No new system was admitted to production, no equilibrium comparison was accepted, and ligand/free-energy exclusions remain exclusions. Additional molecular dynamics has an observable-first plan and restart/storage requirements, not an invented GPU speed or new binding result.
Scoped deliverable verified49. Reconcile current simulation inventory.
Original requirement: Identify completed, failed, excluded, queued and recoverable runs from authoritative outputs. Deliverable: a run ledger with hashes and explicit live-status verification when performed.
Accepted scope: Bounded hashed run inventory; historical records are not current live telemetry.
Record: frontier-20260914-v1 / work package 49. Download the public ledger.
Scoped deliverable verified50. Qualify physical systems.
Original requirement: Check sequence, construct identity, matched WT/mutant preparation, force-field choices and boundary conditions. Deliverable: per-system acceptance/rejection record before production.
Accepted scope: Per-system hold/rejection disposition; zero newly accepted production systems.
Record: frontier-20260914-v1 / work package 50. Download the public ledger.
Scoped deliverable verified51. Audit existing trajectories for interpretable comparisons.
Original requirement: Assess equilibration, stability, independent replicas and effective sampling. Deliverable: usable analysis windows and uncertainty limits; simulation length alone is not convergence.
Accepted scope: Conservative descriptive metric-window disposition; no equilibrium or comparative inference window accepted.
Record: frontier-20260914-v1 / work package 51. Download the public ledger.
Scoped deliverable verified52. Design decision-driven additional MD.
Original requirement: Specify the observable, control, expected decision impact and stopping criterion first. Deliverable: a bounded run plan; actual execution requires readiness and authorized compute scope.
Accepted scope: Observable-first conditional MD plan; no production execution claimed.
Record: frontier-20260914-v1 / work package 52. Download the public ledger.
Scoped deliverable verified53. Resolve ligand/PMF validity issues.
Original requirement: Recheck chemical identity, sampling correlation and the claims attached to historical calculations. Deliverable: valid analyses or explicit exclusions, not a new binding claim from sparse sampling.
Accepted scope: Explicit retained ligand/PMF exclusions; no corrected affinity result.
Record: frontier-20260914-v1 / work package 53. Download the public ledger.
Outcome unmet / preparation complete54. Establish measured resource planning.
Original requirement: Benchmark accepted systems on the actual workstation configuration when available. Deliverable: measured throughput, storage growth and restart plan; do not scale laptop timing into GPU ETAs.
Accepted scope: The intended outcome remains unmet. Useful preparation and alternatives are complete.
Still missing: No accepted-system actual-host throughput/storage-growth/restart benchmark executed.
Reopen when: Intended physical system passes admission and fresh verified RAID budget supports a bounded benchmark.
Usable alternative: Use explicit byte lower bounds and restart/admission checklist; no fabricated GPU rate or silent system substitution.
Record: frontier-20260914-v1 / work package 54. Download the public ledger.
55-58: Tissue electrophysiology
Cell-to-tissue mappings retain explicit unchanged-kinetics and driving-force assumptions. Anatomy remains generic. Six new coarse-grid histories tested one missed pulse with fixed delivery masks. The farther allocation reversed its alternating success pattern without changing its count; the nearer allocation retained six successes. Surrogate-channel comparisons remain conditional.
Scoped deliverable verified55. Bridge cellular scenarios to tissue inputs.
Original requirement: State the gating/conductance assumptions between cell output and tissue models. Deliverable: scenario mappings with no claim of patient calibration.
Accepted scope: Conditional cellular-to-tissue input mapping only, no solver integration or patient calibration.
Record: frontier-20260914-v1 / work package 55. Download the public ledger.
Scoped deliverable verified56. Improve anatomical/model calibration.
Original requirement: Identify appropriate human geometry, regional properties and independent reference data. Deliverable: documented calibration targets or explicit generic-model status.
Accepted scope: Explicit generic-model status and field-level calibration dependencies; no anatomical reconstruction.
Record: frontier-20260914-v1 / work package 56. Download the public ledger.
Scoped deliverable verified57. Test new delivery-heterogeneity questions.
Original requirement: Preserve spatial placement and beat sequence; extend only beyond already completed masks/grid tests. Deliverable: a distinct, verified endpoint comparison, not recycled capture thresholds.
Accepted scope: Six-case coarse-grid missed-stimulus endpoint comparison verified; no continuum convergence, patient calibration or efficacy claim. Optional fine-grid extension not activated.
Record: frontier-20260914-v1 / work package 57. Download the public ledger.
Scoped deliverable verified58. BacNav and S10s comparisons, routes 11A/11B.
Original requirement: Separate surrogate-channel behavior from accessory-protein effects and beta1 context. Deliverable: qualified scenario comparisons; no transfer of source-study efficacy to WT/R104Q by assumption.
Accepted scope: Qualified conditional component/scenario comparison, not new tissue-efficacy simulation.
Record: frontier-20260914-v1 / work package 58. Download the public ledger.
59-62: Proteomics and variant resources
The allele-resolved peptide candidate was checked against two isoforms and a versioned human proteome. Short fragments can lose specificity even when the full peptide is specific in the searched records. Standards and response-factor validation remain experimental. An eight-field R104Q classification-resource comparison found no change, not proof that every database assertion is unchanged.
Scoped deliverable verified59. Consolidate allele-resolved proteomics design.
Original requirement: Review the existing Lys-C/trypsin rationale, peptide uniqueness and isoform numbering. Deliverable: a source-linked candidate/readout brief, not proof of detection or quantitation.
Accepted scope: Only original computational/design/review deliverable; no empirical validation, clinical claim, backup execution or external release.
Record: frontier-20260914-v1 / work package 59. Download the public ledger.
Scoped deliverable verified60. Define proteomics validation needs.
Original requirement: Identify standards, recovery, detection and quantitation requirements. Deliverable: a minimal validation checklist; actual sensitivity and allele counts remain empirical.
Accepted scope: Only original computational/design/review deliverable; no empirical validation, clinical claim, backup execution or external release.
Record: frontier-20260914-v1 / work package 60. Download the public ledger.
Scoped deliverable verified61. Maintain the variant evidence resource.
Original requirement: Add only genuinely new source evidence with release/version tracking. Deliverable: a delta from the completed census, not a redundant census rebuild.
Accepted scope: Bounded eight-field R104Q resource delta, not gene-wide census or all assertion-content comparison.
Record: frontier-20260914-v1 / work package 61. Download the public ledger.
Scoped deliverable verified62. Reconcile public classification claims.
Original requirement: Distinguish functional evidence, mechanism prediction and clinical classification. Deliverable: an evidence/provenance audit; no automatic clinical reclassification from this supplement.
Accepted scope: Scoped clinical/function/mechanism provenance audit; no clinical reclassification.
Record: frontier-20260914-v1 / work package 62. Download the public ledger.
63-68: Manuscripts, figures and communication
A technical supplement, explicit methods, four reproducible scientific figures and an eight-page illustrated report were assembled. Historical overclaims have scoped correction proposals, not silent rewrites. A concise experimental discussion brief and release/version checklist were prepared; the research campaign itself sent no correspondence and made no deposit.
Scoped deliverable verified63. Draft the computational supplement manuscript.
Original requirement: Explain the executed models, benchmark limitations, predictions and falsifiers. Deliverable: a self-contained scientific draft, not competition prose or publication submission.
Accepted scope: Self-contained working scientific draft with methods; not publication or final goal report.
Record: frontier-20260914-v1 / work package 63. Download the public ledger.
Scoped deliverable verified64. Build an intervention-results supplement after execution.
Original requirement: Include adapters, assumptions, robustness maps and tests from A-C. Deliverable: reproducible new results, not promises labeled as findings.
Accepted scope: Executed intervention-results supplement content; final package manifest remains separate.
Record: frontier-20260914-v1 / work package 64. Download the public ledger.
Scoped deliverable verified65. Create audience-specific figures.
Original requirement: Show model disagreement, intervention sensitivity and observation limits from saved arrays. Deliverable: reproducible figures with prediction labels and exact provenance.
Accepted scope: Three reproducible scientific figures with source rows and captions; final PDF remains separate.
Record: frontier-20260914-v1 / work package 65. Download the public ledger.
Scoped deliverable verified66. Audit remaining overbroad historical claims.
Original requirement: Prioritize mechanism closure, MD interpretation and universal route-blocker statements. Deliverable: a scoped correction proposal; preserve published/historical originals and completed corrections.
Accepted scope: Scoped historical overclaim correction proposal, not universal cleanup.
Record: frontier-20260914-v1 / work package 66. Download the public ledger.
Scoped deliverable verified67. Assemble a concise laboratory discussion packet.
Original requirement: Combine the prospective prediction, a discriminating comparison and only essential unresolved configuration questions. Deliverable: an unsent packet; recheck correspondence before any actual outreach.
Accepted scope: Concise unsent computational discussion packet; no finalized experimental agreement implied.
Record: frontier-20260914-v1 / work package 67. Download the public ledger.
Scoped deliverable verified68. Prepare publication/version decisions.
Original requirement: Identify the intended record, version and artifacts needing release. Deliverable: owner-review checklist; local readiness is not authorization to upload, submit or email.
Accepted scope: Owner-review release/version checklist; destination choice and upload remain unapproved.
Record: frontier-20260914-v1 / work package 68. Download the public ledger.
69-74: Integrity, privacy and reproducibility
The original 340 reference-candidate pairs were adjudicated, with 19 navigation changes and a later delta review. Historical simulation language was qualified without changing frozen evidence. Privacy and backup work produced bounded review plans, not a private-content scan or backup certificate. The final local inventory authenticates 466 files; the public subset here has its own manifest.
Scoped deliverable verified69. Adjudicate reference candidates.
Original requirement: Classify the 315 candidates from the latest check as broken, historical, ambiguous or intentional. Deliverable: evidence-backed repair list; do not substitute arbitrary basename matches.
Accepted scope: All 340 captured baseline reference candidates adjudicated; repair list and 19 installed navigation changes, not all links repaired.
Record: frontier-20260914-v1 / work package 69. Download the public ledger.
Scoped deliverable verified70. Resolve three historical simulation-claim defects.
Original requirement: Compare flagged prose with primary arrays and correction ledgers. Deliverable: scoped correction with preserved history, not a blanket numeric rewrite.
Accepted scope: Only original computational/design/review deliverable; no empirical validation, clinical claim, backup execution or external release.
Record: frontier-20260914-v1 / work package 70. Download the public ledger.
Scoped deliverable verified71. Refresh stale navigation and dependency summaries.
Original requirement: Link current work without changing frozen supplement artifacts. Deliverable: a reviewed update plan that distinguishes completed, conditional and source-gated work.
Accepted scope: Reviewed scoped active navigation update; older handoff preserved, no global pointer clearance.
Record: frontier-20260914-v1 / work package 71. Download the public ledger.
Scoped deliverable verified72. Address the additional privacy warning safely.
Original requirement: An existing handoff flags potentially private content outside the restricted home. Deliverable: owner-directed containment review without opening or indexing that content in research scans.
Accepted scope: Owner-directed privacy review preparation without opening restricted content; no containment execution certificate.
Record: frontier-20260914-v1 / work package 72. Download the public ledger.
Scoped deliverable verified73. Define backup verification.
Original requirement: Distinguish saved, committed and independently backed-up artifacts. Deliverable: checksum/restore verification plan with private versus shareable exports kept separate; do not assume the deferred backup happened.
Accepted scope: Only original computational/design/review deliverable; no empirical validation, clinical claim, backup execution or external release.
Record: frontier-20260914-v1 / work package 73. Download the public ledger.
Scoped deliverable verified74. Add reproducibility status for new outputs.
Original requirement: Record input hashes, versions, test outcomes and limitations. Deliverable: an auditable manifest for future intervention results, leaving the completed supplement freeze intact.
Accepted scope: Independently verified 456-file research inventory, source-specific input provenance, 17-script results, tissue campaign manifest, four figures and eight-page visually reviewed PDF. Local reproducibility package only; not biological validation, release approval or independent backup.
Record: frontier-20260914-v1 / work package 74. Download the public ledger.
75-77: Compute operations
A timestamped workstation check and exact-target storage review preceded any proposed large job. Both inspected scientific work directories were retained because safe reclamation was not established. The new tissue campaign used one local CPU worker, not the GPU. No destructive cleanup or automatic expansion of the run was performed.
Scoped deliverable verified75. Verify workstation access and storage before new compute.
Original requirement: Check actual remote availability, free space and intended SATA RAID work locations when execution is requested. Deliverable: current readiness snapshot, not inherited infrastructure claims.
Accepted scope: Timestamped actual-host readiness observation, not present resource reservation.
Record: frontier-20260914-v1 / work package 75. Download the public ledger.
Scoped deliverable verified76. Define bounded compute scheduling.
Original requirement: Separate CPU analysis, GPU MD and model inference; cap concurrency to measured resources. Deliverable: checkpointed queue with observable progress and no unnecessary LLM polling.
Accepted scope: Checkpointed bounded research queue, not an operating scheduler.
Record: frontier-20260914-v1 / work package 76. Download the public ledger.
Scoped deliverable verified77. Audit recoverable storage growth.
Original requirement: Identify generated caches, abandoned temporary work and retained simulation outputs before proposing cleanup. Deliverable: exact-target retention/cleanup plan; no automatic deletion or source-data deduplication.
Accepted scope: Exact-target bounded retention plan: retain both reviewed scientific workdirs; zero proven reclaimable bytes and no cleanup performed.
Record: frontier-20260914-v1 / work package 77. Download the public ledger.
78-80: Experimental and translational ladder
A prospective plan, conditional decision thresholds and mixed-genotype intervention templates are prepared. The actual assay is not locked. The longer-horizon roadmap separates cellular response, delivery, tissue effects, durability and safety. Completing these preparations does not establish any of those biological outcomes.
Outcome unmet / preparation complete78. Test the prospective surface/function predictions.
Original requirement: Use qualified observations and controls to adjudicate the models. At-home deliverable: locked analysis plan; empirical outcome cannot be supplied by simulation.
Accepted scope: The intended outcome remains unmet. Useful preparation and alternatives are complete.
Still missing: Analysis plan is not assay-locked or registered; no experimental test performed.
Reopen when: Actual construct/readout, primary contrast/margin, preparation allocation and QC/uncertainty choices are resolved before relevant outcome access.
Usable alternative: Retain timestamped four-file preparation snapshot, conditional predictions and parameterized decision rules without inventing assay choices.
Record: frontier-20260914-v1 / work package 78. Download the public ledger.
Scoped deliverable verified79. Test intervention response in the relevant mixed genotype.
Original requirement: Assess whether the intervention changes surface and function as forecast, with treated-WT controls. At-home deliverable: decision thresholds and analysis templates, not established rescue.
Accepted scope: Parameterized decision thresholds and templates; numerical assay-specific margins remain to be chosen.
Record: frontier-20260914-v1 / work package 79. Download the public ledger.
Scoped deliverable verified80. Develop the translation evidence ladder.
Original requirement: Identify expression, delivery, tissue effects, durability and safety evidence required beyond cellular results. At-home deliverable: a staged research roadmap with specialist review needs; no self-treatment or patient efficacy inference.
Accepted scope: Only original computational/design/review deliverable; no empirical validation, clinical claim, backup execution or external release.
Record: frontier-20260914-v1 / work package 80. Download the public ledger.
What verification did, and did not, establish
The selected suite passed 17 scripts. Independent dose reconstruction covered all 9,072 rows, with a maximum absolute discrepancy of 3.33e-16 in the tested outputs. Other tests check reference roles, rejection behavior, mathematical limits, preserved bytes and reproduction. A test passing does not validate an unmeasured biological coefficient.
All 17 script records
| Script | Recorded result | Runtime, seconds |
|---|---|---|
| test_foundation.py | Passed | 0.054 |
| test_forecast_v2.py | Passed | 0.038 |
| test_intervention_actions_v2.py | Passed | 0.036 |
| test_monotonicity_v2.py | Passed | 0.050 |
| test_small_bridges.py | Passed | 0.039 |
| test_identifiability_recovery.py | Passed | 0.040 |
| test_rna_flux_ceiling.py | Passed | 0.038 |
| test_record_contract.py | Passed | 0.040 |
| test_activity_bound.py | Passed | 0.040 |
| test_partner_reference_adapter.py | Passed | 0.035 |
| test_quantity_provenance_bridge.py | Passed | 0.310 |
| test_tissue_forecast_adapter.py | Passed | 0.037 |
| verify_v2_analytic_review.py | Passed | 0.037 |
| verify_all_frozen_conversions.py | Passed | 0.240 |
| verify_frozen_preservation.py | Passed | 0.106 |
| verify_pe_lead_reproduction.py | Passed | 0.068 |
| verify_pe_panel_reproduction.py | Passed | 0.208 |
Source SHA-256 identities are in the downloadable record. The table reports prior executed checks; it does not run them in the browser.
The tissue campaign separately checked 72 node maps, six conductance maps and 12 complete state arrays. Each control exactly reproduced its prior first ten beat records. Each paired case had identical states and node maps before the omitted stimulus. The independent grid check reconstructed every coordinate and per-cell conductance, not just the mean.
The six serial local CPU cases took 1,180.500 seconds in total, about 19.68 minutes, and retained 11,931,041 bytes in 116 files. No case hit the declared 600-second limit. No new GPU production simulation was launched. Historical GPU snapshots are not current telemetry.
The original frozen supplement's 108 files retained their recorded hashes. The final local package inventory verified 466 files. This page's smaller public export has a separate manifest because it intentionally omits private material, internal paths and unreviewed artifacts. Hashes establish identity, not scientific truth.
Public-export file identitiesTissue source rows and numerical provenance
The negative results belong in the record
- No new qualified independent partner-function labels emerged from the bounded source update. More scores on the same examples would not repair the failed transfer.
- The named ParSE deposit and publisher packages did not contain the missing per-barcode PSI files. Variant-level summaries do not substitute for junction-level replicate inputs.
- Selected structural records did not establish the target intermolecular interface. This is not proof that no interface exists.
- No new molecular-dynamics system passed admission; existing ligand and free-energy exclusions were retained.
- The storage review established no safely reclaimable amount in the two inspected scientific directories. They were retained.
- Reference-review candidates remain. The research delivery is not a green-repository or independent-backup certificate.
New analysis versus deliberately reused work
New in this campaign: explicit intervention/capacity/activity calculations, combination and selectivity maps, robustness witnesses, noise-aware planning, forecast adapters, expanded editing verification, structural-source qualification and the six-case missed-stimulus comparison. Four new scientific figures communicate those results.
Reused and preserved: the frozen prospective supplement, the failed partner-transfer benchmark, the earlier peak-current/surface non-identifiability results, the existing Lys-C concept and the original regular-pacing mask/refinement studies. Recovering old input sequences and reproducing old tables is a new reproducibility result, not a new biological experiment.
What is usable now, and what reopens the four dependencies
Usable now: conditional intervention comparisons, explicit failure cases, sequence checks, source-authenticated forecast interfaces, prospective decision templates and a finite tissue-history result. There is no need to wait for an experiment to inspect those consequences.
- Original source files, item 41: a legitimate new location or author-supplied workbook/export can reopen empirical reconstruction. Qualified summaries and conditional analysis remain usable.
- Independent benchmark, item 46: new endpoint-compatible labels and a genuinely new transfer question are needed. The failed existing transfer remains a valid result.
- Actual-host MD benchmark, item 54: admit a physical system and verify storage/restart headroom first. No GPU rate is fabricated in the meantime.
- Assay lock, item 78: establish the actual construct, readout, contrast, margin, preparation allocation and quality-control choices before relevant outcomes. The timestamped analysis preparation is not yet an assay-specific lock.
The shortest experimental information request is for an existing sequence-bearing construct map and the feasible qualified readout. That is separate from source-data requests, future tissue calibration, and treatment validation. No private laboratory discussion is published here.
Public data and code
These are exact allowlisted scientific outputs or an explicitly redacted public ledger. The full local workspace is not a public download. No third-party paper attachment, private email, clinical file or system credential is included.
- All 80 work packages and 17 test records
- All 9,072 dose rows
- 2,592 combination comparisons
- 96-group robustness matrix
- Selectivity/capacity boundaries
- Noise-aware ratio comparison
- Public file sizes and SHA-256 hashes
For a fresh local reproduction, save the four Python files linked in Methods together, run the dose generator, then the independent verifier. They require standard-library Python and refuse to overwrite existing output. The verification compares mathematical implementations, not experimental truth.