An agent proposal is structured output from one of the platform's named institutional agents — for example, a candidate signal from the Signal Ingestion Agent, a credibility assessment from the Credibility Analyst Agent, or a draft briefing from the Executive Synthesis Agent. Every proposal lands in a review surface and never reaches subscriber-visible data without an analyst accept. Each accepted proposal is audit-trailed with the agent slug, version, prompt signature, and analyst id.
Why it matters
AI assistance is real on this platform — but it is never authoritative. The contract is: agents propose, analysts dispose. Every public claim has a human signature behind it.
Breakthrough potential is an operator-set flag (and a supporting deterministic score) applied to entities whose combination of cross-source convergence, novel materials/technology, and replication signals merits elevated attention. Every public breakthrough flag has passed operator review. It indicates 'worth watching closely', not 'confirmed'.
Why it matters
It directs scarce reader and analyst attention toward the highest-signal situations while preserving verification discipline — the flag is a watch signal, not a verdict.
Every claim on the platform carries one of six status badges: claimed (someone said it), observed (multiple parties have noted it), disputed (an open challenge exists), verified (independently corroborated by an arms-length source), analyst_interpretation (analyst's reading of the evidence, not evidence itself), and hypothesis (analyst-stated possibility, explicitly speculative). The badge is always visible next to the claim text — never hidden.
Why it matters
Mixing these states is the central failure mode in advanced-energy coverage. The badges are mandatory: every UI surface that renders a claim renders the badge alongside it. There are no exceptions in production.
Commercialization stage is a qualitative, public-safe band derived from commercialization-context activity around a company (pilots, partnerships, manufacturing-readiness language, company↔technology links). It is shown as Pre-commercial, Early-stage, Developing, or Active commercialization. The exact internal score is not exposed publicly; the band is an orientation cue, not a financial or investment judgment.
Why it matters
Knowing roughly how far along a company is — without overclaiming — helps readers calibrate expectations. The platform never states that a product works; it describes observable commercialization activity.
Confidence Score is a 0–100 measure derived from the platform's source-trust engine and the per-signal evidence weight. Inputs include source class (e.g., peer-reviewed journal vs. ecosystem reporting), replication level, independence of the source from the company, and recency. The score reflects how much weight an analyst should assign the signal — it is not a probability that the underlying claim is true.
Why it matters
A score of 75 on a primary-source claim means we are highly confident about who said it and when, not that the claim itself is verified. The platform deliberately keeps origin confidence and truth verification as separate fields — collapsing them would let popular claims masquerade as verified ones.
Convergence is a deterministic 0–100 score combining six factors: how many independent source types mention an entity, commercialization context, institutional participation, recurrence over time, recency, and relationship-graph density. It is rendered as a tier — watch, emerging, strong, or breakout. Convergence measures how much independent evidence points at the same place; it is explicitly NOT a judgment that the underlying technical claim is correct.
Why it matters
In a contested field, the cheapest mistake is treating a single loud source as a trend. Convergence separates durable, multi-source momentum from one-off noise — so attention goes where independent evidence actually accumulates.
Ecosystem reporting refers to third-party publications inside the LENR / advanced-energy ecosystem — sites like E-Cat World, LENR News, and LENR Forum — that report on companies but are not arms-length verifiers. The platform classifies these as source_class = ecosystem_reporting (Phase 14A) and weights their independence at 0.3 (vs. 1.0 for arms-length sources). They contribute to ecosystem_attention metrics but contribute zero to verification scores.
Why it matters
Ecosystem volume can look like validation if you are not looking carefully. The platform's separation of ecosystem_attention from verification is precisely so subscribers can see when a company is generating buzz without generating evidence.
LENR (Low-Energy Nuclear Reactions), also called CMNS (Condensed Matter Nuclear Science), is a research field that investigates anomalous heat, transmutation, and related effects reported in hydrogen-loaded metal lattices and adjacent systems. The underlying physics is debated; the platform tracks LENR-relevant companies, publications, conferences, and validations without endorsing any specific theoretical mechanism.
Why it matters
LENR sits at the intersection of advanced energy and contested science. Coverage requires distinguishing what has been claimed from what has been independently observed and verified. The platform exists precisely because that distinction is rarely made elsewhere.
Mechanism refers to the proposed physical process — for example, hydrogen-metal lattice loading, nickel-hydrogen surface effects, palladium-deuterium electrolysis, plasma-assisted condensed matter reactions, or hydrino formation. The platform's mechanism_families table catalogues recognized mechanism categories, and company_mechanism_links connects companies to the mechanism families they are publicly associated with. The platform tracks mechanism associations without endorsing any specific theoretical model.
Why it matters
Knowing which mechanism family a company claims helps analysts compare like with like, identify cross-company replication, and avoid lumping incompatible approaches together. The platform tracks mechanism-family overlap as one of its convergence-detection signals.
An ontology is a controlled vocabulary plus the relationships between the terms in it. The platform uses several: a mechanism ontology (nickel-hydrogen, palladium-deuterium, hydrino, etc.) at sql/110_mechanism_ontology.sql; a signal-type taxonomy (12 types in lib/intelSignals.ts); a source-class taxonomy (14 classes in lib/sourceTrust.ts including Phase 14A's primary_source_claim and ecosystem_reporting); and a claim-status ladder (claimed / observed / disputed / verified / analyst_interpretation / hypothesis). Ontologies make the platform's outputs comparable across companies and time.
Why it matters
Without an ontology, every analyst writes their own categories and the corpus becomes uncomparable. The platform's ontologies are versioned, auditable, and analyst-led — extensions require a PR, not a silent code change.
A patent is a legal grant by a national or regional patent office (USPTO, EPO, JPO, etc.) giving the holder exclusive rights to an invention for a limited time. The platform tracks LENR-relevant patent filings via the patents table, linked to companies through company_patent_links. Filing or grant of a patent indicates investment in IP and the patent office's procedural acceptance — it is not evidence that the underlying technical claim works in practice.
Why it matters
Patent activity is a leading indicator of commercialization intent and engineering investment. A patent does not mean a technology has been independently validated; it means someone wrote a sufficiently specific document and a patent examiner found it novel and non-obvious by their bar. We surface patents as an activity signal, not a verification signal.
A primary_source_claim is a statement attributable to the covered company or its principals — for example, a Journal of Nuclear Physics post by Andrea Rossi, a Clean Planet press release, a conference talk by a company-affiliated researcher. The platform records these because they are institutionally important (the company's stated position is data) but never elevates them to verified status without an independent source. Phase 14A's discipline forbids automated promotion of primary_source_claim rows to higher confidence states.
Why it matters
A primary-source claim is one of the most common reasons subscribers see a particular activity rate on a company. It tells you what the company is saying about itself — important context, but never proof.
The Signal Feed at /operations/signals is the platform's main operational surface — a unified view of every classified intel signal across the entire covered universe. It composes rows from agent_findings, verification_events, milestones, conference appearances, and analyst observations into one normalized stream. Filterable by signal type, verification state, company, time window, and full-text search. Newest-first by design.
Why it matters
Most platforms in this space publish narrative posts. The Signal Feed publishes structured rows with source URLs, claim status, and confidence per row. That makes it auditable: if you disagree with a row, you can click through to the source and check it yourself.
Source credibility on the platform is a structured profile maintained in lib/sourceTrust.ts. Each source class (peer-reviewed journal, government grant database, conference page, official press release, primary-source claim, ecosystem reporting, etc.) has a profile of six numeric or categorical fields: source_trust_score (0–100), evidence_strength (0–100), publication_quality (0–100), replication_level (none/single/multi), independence_level (0–100), and commercialization_relevance (0–100). Every signal inherits its source's profile, which feeds the confidence score and the ranking model.
Why it matters
The Phase 14A discipline specifically separates 'origin known' (high source_trust_score on a primary_source_claim) from 'truth verified' (verification_events row from an independent source). A company-controlled blog can have high source_trust_score for what it asserts the company said, without that contributing anything to the verification dimension.
Verification on the platform means a verification_events row exists, attached to an arms-length source (a peer-reviewed paper, a regulator filing, an independent laboratory measurement). It is the strongest claim status the platform recognizes. Every verification_events row carries a source URL and an analyst attribution. Importantly, verification is NOT derived from a confidence score — a high-confidence primary-source claim is still claimed, not verified, until an independent source corroborates it.
Why it matters
Treating popular claims as verified is the most common failure mode in advanced-energy coverage. The platform refuses to do this in code: the claim ladder (claimed → observed → verified) is enforced at the data layer, the UI layer, and the agent contract layer.
A personal watchlist is your saved set of companies. The Watchlist Pulse widget on your dashboard shows recent activity across watched companies; alert rules can be scoped to the watchlist. Distinct from the platform-wide editorial watchlist at /companies/watchlist, which lists entities the analyst lead has flagged for priority research attention regardless of any individual subscriber's interest.