Who checks
a claim of
alien life?
When a telescope team announces a possible biosignature, the field argues for years and the public is left guessing. Vyom is the instrument that settles it — one pre-registered, uniform retrieval that returns a reproducible verdict on a published ladder of evidence, for any world in the JWST archive.
How a maybe becomes a mess.
The most famous biosignature claim of the JWST era, as it actually unfolded — claim, counter-claim, refutation. Every entry is a real, cited paper. The lesson is the same each time: with no shared method, the same spectrum yields opposite answers.
- Sep 2023claim
The Hycean claim
Madhusudhan et al. report CH₄ (~5σ) and CO₂ (~3σ) with no NH₃ in K2-18 b's atmosphere, plus a tentative DMS hint — read as a possible ocean world with life.
ApJL 956 L13 - 2024critique
A mini-Neptune fits too
Wogan et al. and Glein show the same chemistry is explained by a warm mini-Neptune with no ocean. The Hycean label is not uniquely supported.
ApJL 963 L7 - Apr 2025claim
“Strongest hints of life”
A MIRI follow-up reports DMS/DMDS at 3σ, ≥10 ppm. Global headlines call it a detection of life — though 3σ is below the field's own discovery bar.
arXiv:2504.12267 - 2025critique
Three independent refutations
Schmidt (60 reductions, 250 retrievals), Welbanks (expanded molecular basis), and Taylor (multiple-testing) each find DMS not robust. CH₄ holds; CO₂ is marginal.
arXiv:2501.18477 · 2505.13407 · 2504.15916 - 2026consensus
The field is stuck
Consensus: the disagreement is methodological — reductions, priors, and molecular bases differ. The same spectrum yields different chemistry. There is no neutral, reproducible verdict. That is the gap Vyom fills.
One pipeline, from spectrum to verdict.
One model, trained once
Amortized neural posterior estimation.
A single conditional normalizing-flow estimator is trained once on 10⁵ physics-based forward simulations (PLATON). After that, it returns a calibrated posterior for any archive spectrum in seconds — no per-target MCMC, no bespoke setup.
One pipeline, applied uniformly
The same priors for every world.
Every public JWST transmission and emission spectrum is run through identical priors, the same forward model, and the same molecular line lists. Heterogeneity — the reason team-by-team results disagree — is removed by construction.
Pre-registered before the verdict
Choices committed before the data is seen.
The analysis plan is registered on OSF before any blind retrieval runs. Priors, model space, and decision thresholds are fixed in advance, so a verdict cannot be reverse-engineered to a desired conclusion.
Calibrated and cross-checked
The posterior is trustworthy, or it's flagged.
Simulation-based calibration certifies coverage; a nested-sampling retrieval (dynesty) cross-checks every critical claim; abiotic-null tests and a strict Bayes-factor threshold gate the final verdict.
Not a yes/no. A rung.
A candidate spectral feature is present in the data above instrument noise.
The feature survives data-reduction, instrument-systematic, and stellar-contamination checks — it is real, not an artifact.
The species is plausibly produced by life, in a habitable context for this planet.
Every known non-biological pathway for the species, on this planet class at this temperature, is ruled out.
The signal is confirmed in a second instrument, wavelength, or epoch — not one visit's quirk.
Independent lines of evidence (companion molecules, disequilibrium pairs) point the same way.
Follow-up leaves no credible non-biological explanation. A claim the community can keep.
Today's loudest claims sit at level 1–2. Vyom's job is to say which rung the evidence actually supports — and prove it.
The second opinion the field lacked.
All six uses →When a biosignature paper lands on a referee's desk, Vyom returns an independent, pre-registered verdict on the same spectrum — a neutral check no single lab can provide on its own work.
Rank every public target by disequilibrium and information content in one pass, so scarce follow-up time goes to the spectra that can actually move a verdict up the ladder.
Run your own target through a neutral, calibrated pipeline before submission — see exactly which of the seven standards your claim passes, and which it doesn't, while there's still time to strengthen it.
When the next "signs of life" headline breaks, point to a reproducible verdict on a published ladder instead of duelling press releases — with the uncertainty stated honestly.
Run a verdict yourself.
Pick a world. Switch reductions and watch the chemistry move. Toggle the corrections that collapsed the DMS headline. Read the verdict on the ladder — live, in your browser.
Open the console