Glossary

Plain-English definitions for ScienceToStartup — Viability Score, Signal Fusion, Buildability Receipt, Paper Pack, Foresight, MCP, llms.txt, AEO Answer Cards, and the rest of the product vocabulary.

111 of 111 termsSource: curated glossary catalogcurated_staticgit_versioned_curated_catalogGlossary API

Scoring

Viability, signal fusion, commercial flags, time-to-MVP.

Capability Index: The machine-readable product graph at /api/capabilities.json. Lists every REST, MCP, and discovery surface.; MCP llms.txt Service Catalog
Commercial Flags: Boolean signals attached to a paper: has_code, has_demo, has_dataset, beats_sota, is_quick_build, is_series_a, is_high_potential.; Viability Score Paper Pack
Composite Score: Generic name for any score that blends multiple sub-scores. Signal Fusion is the canonical composite on ScienceToStartup.; Signal Fusion Signal Fusion Breakdown
Consensus Engine: The multi-model evaluator that compares scoring outputs across LLM providers to keep results consistent.; Viability Score Signal Fusion
One-Liner: A 25-words-or-fewer startup pitch generated for every paper. Names a specific user and a specific action.; Paper Pack Viability Score
Prediction Component: The Foresight-prediction-derived input to Signal Fusion. Weighted by confidence and recency of the frozen prediction.; Signal Fusion Foresight Frozen Prediction
Signal Fusion: The composite score combining viability, prediction, community vote, and GitHub star velocity into one primary ranking number.; Viability Score Foresight Today Rail
Signal Fusion Breakdown: The decomposed view of a Signal Fusion score: viability, prediction, vote, and star velocity components with weights.; Signal Fusion Viability Component Prediction Component
Star Velocity: Daily rate of GitHub star accumulation on the repo associated with a paper. Feeds the Star Velocity Component of Signal Fusion.; Signal Fusion Star Velocity Component
Star Velocity Component: The GitHub-star-derived input to Signal Fusion. Normalised so a brand-new repo does not dominate older mature ones.; Signal Fusion Star Velocity
Time to MVP: Estimated time to a working minimum viable product: under one month, one to three, three to six, or six-plus months.; Paper Pack Buildability Receipt
Viability Breakdown: The decomposed view of a viability score: numeric value, individual commercial flags, and a rationale string.; Viability Score Commercial Flags
Viability Component: The Viability-Score-derived input to Signal Fusion. One of four components that combine into the composite ranking number.; Signal Fusion Viability Score Signal Fusion Breakdown
Viability Score: A 0-10 rating of an AI paper's commercial viability. Higher scores cluster with code, demo, dataset, and beats-SOTA signals.; Signal Fusion Commercial Flags One-Liner
Vote Component: The community-vote-derived input to Signal Fusion. Aggregates trusted-user votes with anti-brigading safeguards.; Signal Fusion Viability Component

Surfaces

The product pages and the canvases behind them.

Build Loop: The build/watch/verify decision surface. Where Signal Canvas asks what is interesting, Build Loop asks should you build it.; Signal Canvas Build Room Launch Pack
Build Loop Dossier: Per-paper Build Loop record holding the verdict (build, watch, verify), the artifacts so far, and the open questions.; Build Loop Launch Pack Build Room
Build Room: An agent-orchestrated session in Build Loop. Multiple specialist agents (wedge, plan, eval, GTM, moat) produce startup artifacts.; Build Loop Launch Pack
Daily Dashboard: Canonical /dashboard surface. Shows today's papers, Today Rail, recent prediction-market activity, and the action rail.; Today Rail Signal Fusion Trends Desk
Discovery Hub: The /discover landing surface. Bundles today's top papers, trending topics, latest articles, and popular questions.; Today Rail Trends Desk Topic
Foresight Backtest Tab: The /foresight backtest tab rendering hit rate, precision-at-k, and the Calibration Curve for each model iteration.; Foresight Backtest Run Calibration Curve
Foresight Flywheel Tab: The /foresight flywheel tab showing publication states (seed_only, outcomes_observed, retrained_published) per model iteration.; Flywheel Foresight Model Iteration
Foresight Predictions Tab: The /foresight tab listing every active frozen prediction with confidence, countdown, and predicted outcome.; Foresight Frozen Prediction Reasoning Factor
Launch Pack: Exportable bundle of Build Room artifacts. The deliverable a founder takes out of Build Loop to start shipping.; Build Loop Build Room
Module Wrapper: The shared React component every workspace, talent, and SEO module wraps with: heading, count chip, error banner, skeleton.; Quiet Canvas Signal Canvas Talent Canvas
Proof Homepage: The root / route. Proof-facing, not dashboard-facing. Shows trust surfaces and platform proof state.; Daily Dashboard Foresight Freshness Ledger
Quiet Canvas: The Claude-language design system: cream surface, hairline borders, ink primary text, amber forbidden.; Module Wrapper Signal Canvas
Signal Canvas: The primary research-synthesis surface. Citation-first search across the corpus with three modes: corpus, paper, workspace.; Research Kernel Paper Pack Workspace
Talent: The expert and builder discovery surface. Seeded from a paper, a workspace, or natural language; returns ranked profiles.; Talent Canvas Warm Intro Path
Talent Canvas: The keystone deep module for /talent. Five slots — search, results, profile, outreach, history — behind one session prop.; Module Wrapper Talent Signal Canvas
Today Rail: The daily-dashboard component that surfaces today's highest-signal papers ranked by Signal Fusion.; Signal Fusion Trends Desk Viability Score
Topic Page: Canonical /topic/{slug} surface for a research area. Carries paper count, average viability, growth rate, top questions.; Topic Per-Topic OG Card Trends Desk
Trends Desk: The narrative trend-analysis surface. Tracks paper volume, scoring trends, and emerging topics with editorial framing.; Topic Today Rail Discovery Hub
Trends Page: The /trends surface — six dropdown-linked pages covering narratives, topics, entities, methods, datasets, and benchmarks.; Trends Desk Topic Corpus Surface
Warm Intro Path: A chain of mutual connections linking the user to a talent candidate. Surfaced on every Talent profile.; Talent Talent Canvas
Workspace: Persistent thesis object owned by an authenticated user. Stores threads, documents, decisions, and run history.; Signal Canvas Build Loop

Agents & MCP

How agents discover and call ScienceToStartup over MCP.

AEO: Answer Engine Optimisation — producing self-contained, schema-anchored content an Answer Engine can quote verbatim.; Answer Engine AEO Answer Card Defined Term Set
Agent Handoff Contract: Typed payload for context transfer between product surfaces. Carries Source Context, REST examples, and MCP examples.; Source Context MCP
Agent Manifest: The structured metadata describing what an agent can do. Schema-versioned so external consumers can rely on it.; MCP Capability Index
Agent Role: A specialist persona inside a Build Room session: product_wedge, build_plan, eval_benchmark, gtm_interviews, moat_risk.; Build Room Build Loop Artifact Type
Answer Engine: An LLM-backed search system (Perplexity, ChatGPT search, Claude, Gemini) that answers a query directly rather than ranking links.; AEO llms.txt AEO Answer Card
Approval Gate: Human-in-the-loop checkpoint for agent actions that require explicit approval before they run.; Sidekick Engine Workspace Operator
Artifact Type: The output type of a Build Room agent: prd_outline, mvp_plan, evaluation_plan, outreach_packet, risk_matrix.; Agent Role Build Room Launch Pack
MCP: Model Context Protocol — the public remote tool surface. Agents call streamable-http MCP tools to query papers, scores, and buildability.; MCP Tool Capability Index llms.txt
MCP Resource: A URI-addressable data source exposed via MCP. Examples: sciencetostartup://contracts/product_graph, buildability://receipt/{id}.; MCP MCP Tool
MCP Tool: A callable action exposed over MCP. Each tool has a typed input schema, output schema, and capability binding.; MCP Capability Index
OpenAI Agents SDK: Open-source agent SDK that consumes ScienceToStartup REST via OpenAPI and ScienceToStartup MCP via streamable HTTP.; MCP MCP Tool
Research Kernel: The backend orchestrator behind Signal Canvas. Plans queries, retrieves sources, compiles answers, and emits SSE streams.; Signal Canvas Source Context
Service Catalog: The canonical JSON manifest of every product surface, route, MCP tool, and handoff contract.; Capability Index MCP
Sidekick Engine: The agent executor inside workspaces. Runs background jobs, executes approved actions, manages Approval Gates.; Workspace Workspace Operator Approval Gate
Source Context: The typed handoff payload between product surfaces. Carries surface, mode, query, paper_ref, thread_id, workspace_id.; Signal Canvas Build Loop Research Kernel
Workspace Operator: The orchestrator agent inside a workspace. Coordinates research threads, document proposals, and run policies.; Workspace Sidekick Engine Approval Gate

Distribution & AEO

llms.txt, DefinedTermSet, FAQPage, OG cards, sitemap pings.

AEO Answer Card: An always-visible, schema-anchored Question/Answer block embedded on a high-traffic page (/, /talent, /buildability, /trends, /dashboard).; FAQ Entry Answer Engine AEO
AI Overview: Google's LLM-generated answer that appears above the classic SERP. Quotes from schema-anchored, visible-on-load content.; Answer Engine AEO Answer Card Perplexity Citation
ClaudeBot: Anthropic's web crawler. Explicitly allowed on /robots.txt for the ScienceToStartup public corpus.; GPTBot PerplexityBot llms.txt
Defined Term Set: The schema.org DefinedTermSet JSON-LD on /resources/glossary. Aggregates every public glossary term for Answer-Engine lift.; AEO Answer Engine llms.txt
FAQ Entry: A typed record (slug, question, answerMarkdown, category, relatedSurfaces) in the public FAQ catalogue.; AEO Answer Card Answer Engine
Fresh Sitemap: /sitemap-fresh.xml lists every URL updated in the last 7 days. Distinct from sitemap-news for shorter recency windows.; Sitemap Index News Sitemap IndexNow
GPTBot: OpenAI's web crawler. ScienceToStartup explicitly allows GPTBot on the public corpus via /robots.txt.; ClaudeBot PerplexityBot llms.txt
IndexNow: Protocol for instantly notifying search engines that a URL has changed. We ping Google and Bing on every sitemap refresh.; Sitemap Index Fresh Sitemap
llms.txt: The agent-discovery file at /llms.txt. Concise plain-text inventory plus inlined Glossary, FAQ, Capabilities, and Deprecations.; Answer Engine Capability Index MCP
News Sitemap: /sitemap-news.xml lists URLs published in the last 48 hours, with stricter Google News schema. Used by Top Stories surfaces.; Sitemap Index Fresh Sitemap IndexNow
Open Graph Card: The 1200x630 social preview rendered by /api/og/{surface}. Per-paper, per-topic, and per-author flavours.; Per-Topic OG Card ScholarlyArticle Envelope
Per-Topic OG Card: The /api/og/topic dynamic Open Graph image. 1200x630 PNG with topic name, paper count, and trend chip.; Topic Discovery Hub
Perplexity Citation: A reference to ScienceToStartup inside a Perplexity answer. Tracked as the canonical AEO outcome metric.; PerplexityBot Answer Engine AEO
PerplexityBot: Perplexity's web crawler. Explicitly allowed on /robots.txt and a primary target for AEO Answer Cards.; GPTBot ClaudeBot Perplexity Citation
RSS Feeds: Per-stream feeds at /papers/feed.xml, /articles/feed.xml, /media/feed.xml, /developers/feed.xml.; Sitemap Index llms.txt
Schema.org JSON-LD: Structured-data vocabulary embedded in pages as JSON-LD. We emit DefinedTerm, FAQPage, ScholarlyArticle, CollectionPage, Question.; AEO Answer Card Defined Term Set ScholarlyArticle Envelope
ScholarlyArticle Envelope: The schema.org ScholarlyArticle plus Person[] (authors) JSON-LD on every /paper/[arxivId] page. Makes a paper citable as primary source.; Paper Pack Answer Engine
Sitemap Index: The root /sitemap.xml file that lists every sub-sitemap. Crawled by Googlebot, Bingbot, GPTBot, and Perplexity.; Fresh Sitemap News Sitemap IndexNow

Data & Corpus

Paper Pack, freshness ledger, topic taxonomy, corpus engine.

Corpus Engine: The full ingestion-enrichment-publish pipeline. Daily Modal jobs at 20:05, 20:30, 21:00 UTC.; Freshness Ledger Viability Score
Corpus Fact Run: Signed assertion about a corpus entity — method exists, dataset is public, benchmark result verified. Carries a provenance signature.; Corpus Surface Signed Receipt
Corpus Ontology: Controlled vocabulary for methods, datasets, benchmarks, and labs. Rejects synonym soup before it leaks into Corpus Surfaces.; Corpus Surface Topic
Corpus Surface: Entity-level aggregation: methods, datasets, benchmarks, labs, opportunity clusters. Each has a slug, edge type, and freshness window.; Corpus Ontology Node Embedding Topic
Freshness Ledger: The per-category staleness record at /api/freshness.json. Three statuses per category: fresh, stale, critical.; Corpus Engine Today Rail
Golden Corpus: Hand-labelled evaluation set with known commercial outcomes. New scoring models must beat the previous generation on it before shipping.; Model Iteration Viability Score
Node Embedding: Graph-structural embeddings for corpus entities — methods, datasets, benchmarks, labs. Used for similarity and clustering.; Corpus Surface Paper Chunker Talent
Paper Chunker: Splits paper full text into overlapping semantic chunks for embedding, retrieval, and citation-anchored answers.; Tier 3 Exhaustive Extraction Research Kernel Node Embedding
Paper Pack: The compact citation-grade envelope for one paper. Title, abstract, score, flags, one-liner, time-to-MVP. Sized for agent context (under 6144 bytes).; Paper Pack v2 Viability Score One-Liner
Paper Pack v2: The current version of the Paper Pack format. Single canonical envelope that agents and MCP tools cite from.; Paper Pack MCP
Tier 1 Light Extraction: The lightest extraction tier: title, abstract, author list, and metadata. Runs on every ingested arXiv paper.; Tiered Extraction Tier 2 Standard Extraction Tier 3 Exhaustive Extraction
Tier 2 Standard Extraction: The middle extraction tier: sections, figures, headline claims. Default for papers that clear a viability threshold.; Tiered Extraction Tier 1 Light Extraction Tier 3 Exhaustive Extraction
Tier 3 Exhaustive Extraction: The deepest extraction tier: full text, tables, formulas, citation graph. Reserved for top-quartile viability papers.; Tiered Extraction Tier 1 Light Extraction Tier 2 Standard Extraction
Tiered Extraction: Three quality tiers of paper extraction: light (title + abstract), standard (sections + figures), exhaustive (full text + tables).; Corpus Engine Paper Pack
Topic: A canonical research area: slug, display name, paper count, average viability, growth rate, proof status, freshness, aliases.; Trends Desk Discovery Hub Per-Topic OG Card

Foresight

Frozen predictions, backtests, flywheel, calibration.

Backtest Run: A historical evaluation of the scoring model against known outcomes. Returns hit rate, precision@k, calibration curve.; Foresight Frozen Prediction
Brier Score: A proper scoring rule for probabilistic predictions. Lower is better. Public Brier receipt is a pending gate for Wave 7.; Foresight Backtest Run
Calibration Curve: A chart of predicted probability against actual frequency. A well-calibrated model sits on the diagonal of the curve.; Foresight Backtest Run Brier Score
Calibration Point: One bucket in the Calibration Curve: predicted probability, actual frequency, and the count of papers in the bucket.; Calibration Curve Foresight
Flywheel: The Foresight self-improvement cycle: predictions, outcomes observed, model retrained, better predictions, repeat.; Foresight Backtest Run
Foresight: The verifiable-prediction surface. Public ledger of frozen predictions with backtests, reasoning chain, and a self-improving flywheel.; Frozen Prediction Backtest Run Signal Fusion
Foresight Batch: A cohort of frozen predictions minted together. Carries batch id, model version, formula version, and a lineage hash.; Frozen Prediction Foresight Model Iteration
Foresight Ledger: The public, append-only record of every frozen prediction batch with hit-rate, Brier, and calibration receipts.; Foresight Frozen Prediction Brier Score
Frozen Prediction: A Foresight prediction whose scores were locked at mint time and cannot change. The public ledger is built from these.; Foresight Backtest Run
Hit Rate: The fraction of frozen predictions whose top-ranked outcome matched what actually happened by the countdown date.; Foresight Backtest Run Brier Score
Model Iteration: One generation of the scoring or prediction model. Each iteration carries a version, training data hash, and Golden Corpus benchmark.; Flywheel Golden Corpus Foresight
Predicted Outcome: The category Foresight assigns to a paper: startup, oss_traction, acquisition, adoption, or none.; Foresight Frozen Prediction
Reasoning Factor: One factor in a Foresight reasoning chain. Carries name, category, weight, contribution, evidence, source reference.; Foresight Frozen Prediction

Buildability

Receipts, build passports, scaffolds, proof lab, waves.

Adversarial Divergence: Judge-run comparison test that measures disagreement between independent evaluators on the same Buildability Receipt.; Buildability Wave Buildability Receipt Signed Receipt
Build Passport: Reproducible build artifact bundle for a paper: pinned Dockerfile, RUN.sh, EXPECTED_OUTPUT.json, SBOM, cost estimate.; Buildability Receipt Proof Lab
Buildability Check: One typed check inside a Buildability Receipt. Status (pass, fail, skip, pending), evidence, and a blocker flag.; Buildability Receipt Build Passport
Buildability Receipt: The receipt for a paper's buildability. Carries status, checks, evidence set, holdout summary, scaffold contract, external gates.; Buildability Check Build Passport Buildability Wave
Buildability Wave: Progressive buildability assessment phases, wave_0 through wave_7. Each wave adds evidence requirements and external gates.; Buildability Receipt Brier Score
External Gate: A precondition that must clear before a Buildability Wave can complete. Common gates: founder signoff, public Brier receipt, divergence.; Buildability Wave Signed Receipt Verify Receipt
Hindsight Calibration: The retrospective check that compares issued Buildability Receipts to the build outcomes that actually happened.; Buildability Wave Hindsight Public Ledger Brier Score
Hindsight Public Ledger: Public record of Buildability predictions vs. real-world outcomes. Three states: seed_only, outcomes_observed, retrained_published.; Hindsight Calibration Buildability Wave Model Iteration
Proof Lab: The service that runs reproducibility checks against paper code. Budget-capped at seven dollars a month, twenty-three cents a run.; Build Passport Buildability Check
Scaffold Contract: The build scaffold for a paper: surface (Dockerfile, notebook, script), status, source kind, artifacts.; Buildability Receipt Build Passport
Signed Receipt: Any platform receipt with an ed25519 signature and a key id. Verifiable via the public verify-receipt endpoint.; Buildability Receipt Verify Receipt External Gate
Verify Receipt: Public endpoint at /api/buildability/verify-receipt. Performs fail-closed ed25519 verification of a receipt.; Signed Receipt Buildability Receipt
Wave Completion Receipt: Signed receipt certifying a Buildability Wave is complete. Carries status, requirements, founder signoff, and lineage hash.; Buildability Wave Signed Receipt Verify Receipt