SMOS — Scientific Memory OS

Goal-conditioned trajectory optimization for research.

SMOS converts your research objective into goal coordinates, scores every paper in your corpus against them, and identifies the complementary papers that advance your trajectory — not just the similar ones.

How it works

Full pipeline. One workspace.

01 Anchor
Set your goal
Upload a Master Paper + objective sentence. SMOS extracts Master Claims — your goal coordinates.
▶
02 Ingest
Build your corpus
Upload PDFs or search by DOI. Full-text extraction (+60% claims vs abstract). Auto-dedup, iDOI stamping.
▶
03 Score
Align every paper
Project Alignment Score (PAS) — cosine similarity of each paper's claims vs your goal coordinates. Sovereign Threshold filters the corpus.
▶
04 Expand
Find what's missing
Complementarity engine generates 5 paper slots per anchor: missing axis, mechanism gap, bridge, contradiction, recent primary.
▶
05 Certify
Protect IP
Shadow Ledger records certified failures (Negative iDOI). Immutable hash-chained ledger provides legally defensible prior art.
Galaxy — Cluster Map

See your entire knowledge landscape in one view.

TF-IDF + PCA 2D projection of your full corpus. Every paper is a node; KMeans clusters surface the conceptual territories. Gaps in the map are where research is missing — not just where you haven't looked.

Cluster-level epistemic profiles (7 types: consensus, controversy, uncertainty…)
Scientific Atlas: bibliographic coupling → Voronoi family territories
Bridge detection: Logic Registry scores claims that span clusters
Galaxy — live preview
● Cluster A ● Cluster B ● Cluster C
GPS — Goal Progress Signal

Measure how close you are to your research objective.

GPS computes the A* path from your current corpus state toward the Master Claims. It surfaces the papers and claims that have the highest marginal contribution to your research goal — not just the highest signal scores.

Project trajectory over time — see progress or drift
Per-paper marginal contribution to the goal vector
Alerts when temporal drift erodes claim certainty
GPS — trajectory toward goal
Goal
★
PAS
72
Highest marginal contributors
Johnson et al. 2023 · Δ +8.4 pts
Wang et al. 2022 · Δ +6.1 pts
Patel et al. 2024 · Δ +4.7 pts
Shadow Ledger — IP Defense

Certified failures are a competitive moat.

Most labs discard negative results. SMOS preserves them as Negative iDOI stamps — anonymized by mechanism, immutably hash-chained, legally defensible. One certified failure can invalidate a competitor's patent claim.

Negative iDOI format: NEG-{ORG}-{YEAR}-{COUNT}
SHA-256 hash-chained ledger — every entry is tamper-evident
Ghost Resonance: TF-IDF scan alerts when live claims echo known failures
Shadow Ledger — recent entries
NEG-ATL-2026-031 Oxidative stress — lipid nanoparticle failure mode A
NEG-ATL-2026-030 Tau aggregation — inhibitor mechanism B
NEG-ATL-2026-029 Blood-brain barrier — crossing protocol C
⚠ Ghost Resonance alert Claim #1247 echoes NEG-ATL-2026-031 (similarity 0.83)
Benchmarks

Numbers from the Alzheimer's corpus — live.

329
Papers in live corpus
3,414
Claims extracted & scored
+60%
More claims vs abstract-only
+32 pp
Quantitative signal gain (full-text)
Competitive landscape

SMOS finds what's missing. Others find more of the same.

Capability SMOS Semantic Scholar Elicit Consensus
Goal-conditioned alignment (PAS) ✓ ✕ ✕ ✕
Complementarity engine (finds gaps) ✓ ✕ ✕ ✕
Certified failure library (Shadow Ledger) ✓ ✕ ✕ ✕
Full-text claim extraction ✓ ◐ ✓ ✕
Contradiction detection (Diamonds) ✓ ✕ ◐ ◐
Immutable audit trail (Ledger + iDOI) ✓ ✕ ✕ ✕
Multi-tenant workspace (teams) ✓ ✕ ◐ ✕
Next step

See SMOS on your corpus.

We'll run a live demo on a corpus you bring — or on the Alzheimer's corpus if you'd like to see a deep example.

Request a demo → Meet the team