Scientific publishing · evidence first
A living scientific library, built from traceable evidence.
Monographs assembled from published research. Every assertion carries its source, its exact locator and its epistemic status — and the reader places them beside the text, never merged with the synthesis.
What is actually here
measured, not claimedSource: catalog_projection (proiezione locale degli eventi firmati dell'Engine). Counts are read from the rows, never estimated.
Disciplines
declared navigation facets1 edition
1 edition
1 edition
0 editions
1 edition
0 editions
0 editions
0 editions
Facets are a declared editorial taxonomy; a volume appears under a discipline only when one of its recorded topics matches a declared rule. 0 editions are currently unclassified — that is shown as a fact, not smoothed over.
Latest volumes
from the live catalogue- 01Exoplanet Atmospheric Characterization
Retrieval Methods and Open Questions
- 02Exoplanet Atmospheric Retrieval: Spectroscopy, Bayesian Inference and Physical Degeneracies
A Traceable, Evidence-Anchored Synthesis
- 03Dark Matter: Direct Detection and Astrophysical Evidence
A Traceable, Evidence-Anchored Synthesis
- 04CRISPR Gene Editing: Mechanisms, Delivery and Off-Target Effects
A Traceable, Evidence-Anchored Synthesis
Evidence Lens
In the reader, any assertion is a link: clicking it opens the exact source — DOI, authors, year, the paragraph or section it came from, the evidence type and the epistemic status the Engine recorded. If a claim has no verifiable evidence in the database, the reader says so; it never fills the gap with a generated citation.
Full-text versus abstract-only sources are distinguished, and retractions are reported as measured facts from the connectors.
what a claim shows
Assertion in the text → source · locator · evidence type · status
No claim records available (engine responded 404). The panel renders only fields present in the Engine's records, and missing evidence is flagged as missing.
Reading, not scrolling
the academic readerIn the product today:
- Per-chapter delivery. One chapter at a time from the pinned structure contract — the whole volume is never loaded in the browser.
- Typography for long text. Serif reading measure (66ch), size and light/dark theme under reader control.
- Equations. KaTeX with strict parsing; the validation status recorded by the Engine is shown as recorded.
- Data figures. Deterministic SVG, hash-checked against the declared hash, never decorative.
- Evidence panel. Each claim opens its source, locator, evidence type and epistemic status.
- Bibliography. Real registry records, with access scope and measured retractions.
- Bookmarks, notes, resume. Stored on your account and verified server-side.
- Search inside the volume. Only recorded blocks are matched.
- Keyboard navigation. J/K between chapters; skip link and focus rings.
- Preferences on your account. Size and theme are read from the account record at page load — no value is presented as saved when no record exists.
- Translated versions. Only jobs the Engine has completed are listed, with the status exactly as recorded — a recorded request is never presented as a readable version.
Declared as planned (not in the product):
- Editions produced before M7 P7 have no structure asset: they render as published, and the Portal declares that fallback instead of reconstructing content.
Open a reader — requires an account and an entitlement.
Access and plans
Free tier: register and choose up to three complete volumes to read online. The count is enforced server-side and recorded as an operation, so concurrent requests cannot overdraw it.
Translations are a separate, asynchronous request — not part of the reading subscription, and priced by declared policy rather than assumed.
Subscription
pricing not configured
No payment is processed in this build: the payment adapter runs in test mode and refuses a live mode without explicit activation.
Cross-disciplinary paths
No dossier records available. The Engine's rules engine produces these; nothing here is generated to fill the space.
What we do, and what we do not claim
- Declared method. Discovery, retrieval and writing run under recorded gates; an independent automated review engine checks claims against sources.
- No peer-review claim. Work is human-reviewed only when a named human review actually happened; statuses distinguish automated review from human review.
- No invented numbers. Accreditations, usage statistics and "impact" figures are not displayed unless measured from real records.
- Image generation is a presentation layer. Cover art is produced by the Engine's image module with declared provenance and manual approval; data figures remain deterministic. Currently running in mock mode.