Updated August 29, 2026

Evidence-guided science

Scientific claims with supporting evidence, scope, uncertainty, and testable next steps.

Featured insights

Does the wavefunction exist?

The wavefunction is a successful mathematical state-description; whether it represents physical reality remains interpretation-dependent because competing views can make identical predictions.

Dark energy and accelerated expansion

Accelerated cosmic expansion is well measured, whereas the physical cause remains unknown; current evidence still favors ΛCDM despite unresolved tensions.

Urolithin A and human longevity

Urolithin A has plausible cellular and animal mechanisms, but sustained human exposure has not yet been shown to extend lifespan or reduce age-related mortality.

Flagellar motility, energy, and fitness

Motility can improve navigation while consuming energy, so environment-dependent mixtures of motile states are a plausible but unconfirmed evolutionary strategy.

Spatial competition and resident diversity

Localized resource use may suppress invaders while preserving resident niches, but direct evidence separating depletion from broader competition remains limited.

AI in clinical decision-making

Clinical AI raises accountability, privacy, equity, and over-trust risks, especially when explanations do not reliably reflect model sensitivity or failure modes.

AFB1 genotoxicity

AFB1 requires metabolic activation to form DNA adducts; mutation risk depends on lesion persistence, repair, replication, and clonal selection rather than adduct count alone.

Why temperature inversions suppress convection

Buoyancy compares air parcels at the same pressure; warm air above colder surface air is already elevated, and the stable temperature gradient suppresses further rising.

Extremophile biology

Extremophile survival repeatedly involves macromolecular stability and energy conservation, while proposed DNA-physical-property signals remain correlational.

Phenotypic clusters in small fiber neuropathy

Clusters may reflect different targets—distal axons, sensory ganglia, or pain-processing networks—but loss and gain of sensory function can coexist.

Paper reviews

Pathology foundation-model atlases

Patient-level controls show that apparent atlas performance can reflect representation signal, tissue composition, technical effects, and decoder choices; independent paired validation is still needed.

Oligodendroglia and multiple-sclerosis risk variants

Multimodal functional-genomics data connect selected variants to oligodendroglial regulation, but engineered perturbations and limited cellular models prevent claims of primary disease causation.

Large language models for malaria drug discovery

TxGemma-9B leads broad separation while LlaSMol-Mistral concentrates hits at the top of the ranking; retrospective assays and limited replication leave prospective value unconfirmed.

Quantitative metagenomics for antibiotic resistance

Sequencing depth alone cannot solve low-abundance stochasticity and homologous read assignment; extraction controls, biological replication, and target-specific validation remain important.

Automation of systematic reviews with language models

Screening and extraction results support supervised acceleration, not autonomous review, because rare or context-dependent errors may alter final evidence synthesis.

Ribosome-collision detection from rRNA fragments

The method repurposes Ribo-seq reads as a discovery screen across organisms, but RNase and library-protocol effects mean direct collision assays remain necessary.

Cellular activity scales in epithelial remodeling

Velocity-correlation shape may distinguish modeled remodeling mechanisms, but the conclusion is computational and lacks validation in tissues with independently known force generation.

Episodic memory for goal-directed planning

Explicit trajectory memories support adaptable, inspectable planning in synthetic graphs, while claims about biological implementation, energy use, and broad autonomy remain provisional.

Phenotyping replication in sweet sorghum

Improved phenotype precision often outperformed larger training populations in the tested Haitian population, but leakage-sensitive validation and economic trade-offs remain unresolved.

Entropy production and computational accuracy

The proposed bound links lag error to dissipation aligned with slow output modes, but evidence is mainly near-equilibrium simulation in reversible or locally reversible models.

Rhamnogalacturonan-II oligosaccharide tools

Engineered microbial pathways expand structural glycan tools, while intact-substrate transport and cross-strain uptake remain indirect or incompletely characterized.

Arabidopsis MACP2 and starvation autophagy

Genetic, imaging, and flux data support MACP2 involvement in ROS-linked autophagy, but pore formation and direct hydrogen-peroxide transport are not established.

Climate-dependent ammonia emissions from poultry

The process-based model offers pathway-resolved global estimates, but simplified management and limited direct validation make results scenario-capable estimates rather than measurements.

Ultraviolet radiation and skin cancer

The review coherently links UV damage, repair, immunity, and clonal selection, but narrative-review methods limit quantitative risk interpretation.

The Evidence Layer for Scientific AI

Ask questions across claims linked to experiments, exact results, limitations, and sources.Know what the science actually supports before you trust the answer.

Explore by Goal

Evidence-guided science

Featured insights

Paper reviews

Experimental ideas

Ask a Follow-Up

Stay current without chasing every paper.

My BGPT

Trending