Every claim below is stated with its supporting evidence, scope, and the uncertainty BGPT identified. Click any title to inspect the full evidence record.
Content updated September 11, 2026
The wavefunction is securely established as a mathematical state-description that generates experimental probabilities, while whether it is a physical object, a state of knowledge, or an emergent description remains unresolved.
Accelerated cosmic expansion is securely measured and current evidence favors a constant-density cosmological constant, but the underlying physical mechanism remains unidentified and some independent measurements are in tension with the standard model.
A warm inversion "lid" is usually already above colder, denser air, so the standard buoyancy argument that denser air above should rise does not apply because density differences at different pressures do not create buoyancy at the same pressure.
A single engineered conjugative plasmid carrying entry-exclusion genes blocked over 99.9% of resistance-plasmid acquisition in vitro and prevented gut colonization in mice, but results rest on small groups (n=3), one mouse strain, and no human-relevant validation.
Entry-exclusion acts as an evolutionarily pre-tested anti-conjugation firewall in laboratory and mouse systems, yet escape-mutant dynamics and coverage of the broader clinical plasmid pool remain untested.
Reducing plasmid maintenance and transfer is biologically plausible in simplified laboratory systems but not clinically established, since within-host adaptation, biofilms, and alternative mobile elements could preserve resistance even after apparent curing.
A contained biofilm design can track transfer, loss, replacement, resistance phenotype, and fitness simultaneously, with the IncX1-versus-IncU conjugation-frequency gap suggesting the bottleneck is mating-pair formation, maintenance cost, or spatial accessibility.
Building and operating flagella can consume a substantial fraction of cellular energy, so selection favors motility only where its benefit exceeds its cost, but the hypothesis that heterogeneous motility distributions reduce population-wide energy expenditure is testable rather than established.
Environment-responsive distributions of motility states are a plausible explanation for why populations keep exploratory capacity while cutting energy use, but direct measurements of within-population heterogeneity and its costs are lacking.
Localized resource use can suppress a neighboring population while leaving other patches and taxa relatively unaffected, but the design principle that spatially narrow competition preserves resident diversity is a testable inference the cited studies do not directly measure.
Roughly 61% of the germline genome is discarded from somatic cells at the 12-cell embryonic stage at precise 29-bp palindromic break motifs, supported by strong genomic and cytological evidence, though the causality of the motif and the nuclease involved remain unresolved.
The A. rhodense discovery supports the idea that eliminated regions become release valves for repeat amplification, but this evolutionary interpretation is inferred rather than directly demonstrated.
CYP450 oxidation converts AFB1 into a reactive epoxide that forms DNA adducts, with the more persistent FAPy adduct driving mutations, while oxidative stress is secondary rather than the established initiating mechanism.
Cell types diversify through specialization, regulatory control, and physical organization into tissues and circuits, but the evidence spans only selected systems and does not establish a single universal evolutionary pathway.
An engineered RNA-allosteric switch controlling protein degradation enables real-time RNA imaging in whole plants, a genuinely new capability, though only two of twelve switch designs worked and independent replication is absent.
In the Drosophila follicular epithelium, TGF-beta signaling strengthens tricellular adhesion by enriching E-cadherin even when Myosin II contractility is suppressed, but the precise redistribution-versus-synthesis mechanism is not proven.
Urolithin A has a biologically plausible but unproven route to supporting healthy aging, with evidence drawn mainly from mechanistic studies, cell models, and mice rather than randomized human survival data.
A bottom-up accounting model estimates roughly 41 kcal/day of tissue-specific cost, about ten times smaller than standard surplus prescriptions, but the estimate contains zero primary data and its dominant parameter derives from infants and animals.
Dual agonists outperform GLP-1 drugs on HbA1c and weight in trial data, but the reviewed systematic evidence is methodologically weak, and independent network meta-analysis corroborates the efficacy direction with far more rigor.
A large US cross-sectional study (N=21,851) links higher PM2.5 to sperm DNA fragmentation, but the design, ZIP-level exposure proxies, and missing smoking and BMI data permit only cautious non-causal inference, and the age effect is far larger than the pollution effect.
Thirty-three Sicilian centenarians showed durable 1918 H1N1 antibody memory but zero SARS-CoV-2 cross-reactivity, meaning any pandemic-era resilience would need to act through mechanisms other than antibody cross-protection, in a cohort too rare to replicate prospectively.
Adult-restored FMR1 expression suppressed audiogenic seizures in mice, implying a circuit defect assumed to require early intervention is dynamically reversible, but the findings are limited to male mice, small groups, and one auditory phenotype.
A specific EBV BRRF2 epitope acts as an antigenic center for CNS-directed cross-reactive antibodies in roughly 8% of MS patients with high serum-test specificity, but causality, T-cell involvement, and strain variation remain open questions.
UV exposure creates DNA damage while repair capacity, immune editing, and clonal competition determine which damaged cells persist, a coherent framework drawn from a narrative review that conducted no original dataset or systematic search.
In anesthetized rats, local and regional brain dynamics largely reverse between loss and recovery of consciousness while global network trajectories do not, suggesting emergence is an active reorganization rather than a simple reversal, but the study covers only male rats and one anesthetic.
A dual-neurotransmitter model in which dopamine drives reward seeking while noradrenergic amygdala signaling drives loss aversion is plausible and testable, but it rests on correlational imaging and small pharmacological studies rather than direct causal tests.
A computational model formalizing intimacy as a graded cost of vulnerable actions tracks human belief updates at r=0.987-0.991 across six preregistered experiments, though cultural generality and multiple relationship dimensions remain untested.
A proposed von Economo neuron fatigue hypothesis links mission duration to emotion-recognition decline, supported by convergent but indirect evidence from public datasets and one human observation, so the central causal chain remains inferred.
Carollia bats phase-lock echolocation calls to low-noise phases of 4-Hz modulated noise in the lab and wild, with causal cortical evidence, though modest sample sizes and a single species and masker frequency limit generalization.
LTP stimuli drive Arc capsid assembly and vesicle release that suppresses AMPA receptors in neighboring neurons, demonstrated in cultured rodent neurons with rigorous methods but no in vivo functional validation.
Nanoconfinement within a solid-state nanopore stabilizes a liquid pathway at -5 to -20 °C and stretches DNA translocation about 400-fold, but the liquid state is inferred electrically rather than visualized and the mechanism remains hypothetical.
Hydrothermal reaction networks produced autocatalytic peptides and vesicles, but the central universal-theory claim rests on analogy rather than direct test, with only two networks examined and no statistical comparisons reported.
Methane ebullition rises with falling atmospheric pressure across 164 mountain waters, consistent with a quantitative degas model, but the evidence is correlational and geographically biased toward the Tibetan Plateau.
A quality-factor screening framework offers a low-cost falsifiable triage for frequency-multiplexed biological information carriers, though its conclusions depend on proxy coherence times and apply only to frequency-based codes.
A leakage-controlled benchmark shows reported codon-LM advantages in synonymous-variant prediction collapse or reverse under gene-held-out splits, a memorization canary outperforms every neural model, and all datasets and code are publicly released.
A pretrained latent-diffusion framework improves reconstruction and clustering of sparse spatial assays across multiple cancers, though its generality depends on similarity to a somewhat imbalanced atlas and inferred genes should not be treated as measured evidence.
Fine-tuned models substantially improve antimalarial virtual screening over tested baselines, with the best model depending on whether global ranking or top-1% enrichment matters, but the evaluation is retrospective with only two fine-tuning replicates and no prospective validation.
A multi-model workflow improves screening and extraction efficiency at review scale, supporting a validated human-supervised process rather than autonomous reviews, with generalizability and error propagation into final conclusions incompletely established.
A design-stage governance prototype converts traceability and escalation rules into testable software invariants, but its biological hazard detection remains proxy-based and effectiveness in deployed workflows is unestablished.
Durable AI memory records should carry type, source, confidence, and permitted use, surviving only when independent evidence supports them, a design principle argued from both cognitive and engineered memory evidence rather than demonstrated at scale.
A foundation model trained on 23 million spatial microenvironments forward-simulates tissue responses and yielded one validated drug combination, though only one designed combination was tested in one in-vitro assay and model code is not public.
A clearly annotated ACS Nano TOC graphic is scientifically defensible as a design hypothesis, with the endosomal-release step best supported by siRNA evidence while tumor accumulation, receptor identity, and apoptosis remain formulation- and model-specific claims.
Know what changed, what holds up, and what remains uncertain. Every Friday. No ads.
Explore by Goal
Ask a Follow-Up
Trending