Answers for teams
trying to understand games.

Latent Spark is Iridae’s calibrated belief map of game design and market space. It turns public industry evidence and private development signals into structured belief states, so teams and systems can understand released games, unfinished concepts, and plausible futures in the same atlas.

It does not treat games as fixed tags, static records, or point embeddings. Each game is represented as an uncertain belief state, so similarity, comps, positioning, forecasts, plans, and experiments all inherit the same calibrated view of the game.

No. Latent Spark can stand alone as a belief atlas, comp engine, similarity layer, and game-understanding API. Anchored Horizon, Patient Cartographer, and Subtle Beacon deepen the loop when connected.

Latent Spark is for anyone who needs systems to understand games: marketers reading the market, designers exploring design space, product and strategy leads evaluating concepts, portfolio teams comparing opportunities, and internal-tool builders who need game intelligence as infrastructure.

Internal tools can call Latent Spark as a game-understanding layer: retrieve comps, classify mechanics, inspect belief states, query similarity, explain clusters, or place internal projects inside the broader market map.

Latent Spark ingests public industry evidence such as store pages, catalogs, screenshots, reviews, tags, community discussion, and market signals. The goal is not to trust any single source, but to accumulate evidence across many sources.

The belief state stays uncertain. That is intentional. Latent Spark would rather show a wide, honest belief than give teams false precision from sparse or conflicting data.

No. Public industry evidence forms the shared atlas. Tenant streams such as Slack, Jira, GitHub, uploads, telemetry, and surveys are private to the tenant and scoped through tenant-isolated storage, retrieval, and API boundaries.

Teams can ask what games are near a concept, which comps are defensible, what positioning is crowded, what adjacencies exist, how a game differs from its neighborhood, and which market assumptions are still weak.

Teams can ask what mechanics and attributes define a concept, what happens if the game moves in a different direction, which design spaces are adjacent, and what existing games resemble the thing the team is becoming.

Yes. That is one of its core advantages. An unfinished concept can be represented as an uncertain belief state and projected into the same atlas as released games, so teams can reason about plausible futures before the market has direct data on the title.

Yes. Latent Spark can support queries like “what if this became more co-op,” “what if we leaned more premium,” “what if this moved toward roguelike structure,” or “what games would this resemble if we changed the core loop?”

It can help identify underoccupied or weakly served regions of design and market space. It also shows uncertainty, which matters: some gaps are opportunities, and some are simply places where available evidence is thin.

A belief state is a structured representation of a game that includes both what the system believes and how confident it is. Latent Spark tracks mechanics, identity, sentiment, commercial shape, and psychographics as distributions rather than single fixed values.

Tags are useful evidence, but they are not enough. They are often incomplete, inconsistent, overbroad, or commercially shaped. Latent Spark can use tags as inputs while still maintaining a richer belief state that includes confidence, conflict, provenance, and similarity structure.

Embeddings can retrieve similar text or content, but they usually hide uncertainty and make it hard to inspect why something is similar. Latent Spark combines residual identity embeddings with interpretable mechanics and attributes, so teams can see both the neighborhood and the reasons.

An unfinished game starts as an uncertain position inside the industry atlas. As private evidence arrives from documents, prototypes, planning tools, playtests, telemetry, or team discussion, its belief state sharpens.

The game is not blank. Latent Spark starts from population priors derived from the industry corpus, then keeps uncertainty wide until tenant-specific evidence arrives. That means systems can operate from day one without pretending to know more than they do.

Latent Spark records current belief. Anchored Horizon forecasts from it. Patient Cartographer plans through it. Subtle Beacon reduces the uncertainties that matter. New evidence returns to Latent Spark, improving the belief state for the next decision.

Yes. Standalone, Latent Spark is a game intelligence layer for comps, positioning, design-space exploration, market mapping, internal tooling, and concept evaluation.

Yes. Latent Spark is delivered as hosted multi-tenant SaaS with tenant isolation, observability, lineage, and API surfaces for products, analysts, and agents. Deployment variants can be discussed with Iridae.

A title, concept, or catalog entry is enough to begin. Industry priors provide an initial uncertain position; connected private evidence sharpens the belief state over time.

Tenant evidence is scoped privately across ingest, storage, cache, retrieval, and downstream access. Shared industry evidence remains shared-read; private development evidence does not become part of another tenant’s atlas.