[ { "_derived": { "prediction_status": "open", "priority": 3.178 }, "applies": [ "llm-agent-societies", "cicero-strategic-dialogue", "constrained-decoding" ], "design_sketch": "The player appoints an envoy with a mandate (export grain, import steel, never sell land). Envoys meet other envoys in a shared hub, negotiate deals constrained by mandates, and report back with transcripts and outcomes the player can ratify.", "falsifiable_signal": { "criteria": "Feature first shipped after 2026-10-24 in a title with at least 1,000 Steam user reviews or a console release; developer material (store page, patch notes, documentation, or talk) states that an AI agent acts for an absent player in trades or negotiations with other players.", "prediction": "A game will let each player's AI representative trade or negotiate with other players' worlds while those players are offline.", "resolve_by": "2029-06-30" }, "feasibility": 3, "horizon": "1-2y", "id": "agent-ambassadors-async", "impact": 3, "involves": [ "async-shared-region", "trade-routes", "faction-diplomacy", "contracts-orders", "natural-language-interaction" ], "key_risks": [ "Agents exploited by adversarial players.", "Mandate violations create disputes.", "Server cost of agent negotiation." ], "name": "Agent ambassadors between player worlds", "novelty": 4, "opportunity_type": "new-mechanic", "precedent": [ "simcity-2013", "animal-crossing-new-horizons", "eve-online", "offworld-trading-company" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Let each player's simulation send language-agent representatives that trade and negotiate with other players' worlds while they are offline.", "thesis": "Asynchronous shared regions (SimCity 2013) and visits (Animal Crossing) connect players thinly because the other player is absent. Agents that represent a player's interests, constrained by the player's standing instructions, can conduct trade and diplomacy between worlds asynchronously, turning a set of single-player games into a living network.", "type": "opportunity", "validation_experiment": "Run 100 simulated player worlds with envoy agents under mandates for 30 days; measure mandate violations, trade volume, and player-rated fairness in a pilot. Success: zero hard mandate violations and positive fairness ratings." }, { "_derived": { "prediction_status": "open", "priority": 2.5 }, "applies": [ "human-behavior-simulation-models", "mean-field-games", "llm-agent-societies" ], "design_sketch": "Each market has a population of agent traders with budgets, risk profiles, and information limits matching human traders, operating under the same order-book rules. Their share of volume is capped and published. Designers tune agent populations to target liquidity and volatility bands.", "falsifiable_signal": { "criteria": "Announced or shipped after 2026-10-24 in a persistent online game with a player-driven market; developer material states that AI market participants were trained or calibrated on player trade data. NPC vendors with fixed prices do not count.", "prediction": "A persistent online game will add AI traders to its player-driven market, with their behavior fitted to observed player trading.", "resolve_by": "2029-06-30" }, "feasibility": 3, "horizon": "1-2y", "id": "ai-liquidity-in-player-markets", "impact": 2, "involves": [ "player-driven-market", "persistent-world", "auctions", "supply-demand-pricing" ], "key_risks": [ "Players perceive bots as unfair competition.", "Agents can be exploited for arbitrage.", "Calibration drift as the human population changes." ], "name": "AI liquidity in player markets", "novelty": 3, "opportunity_type": "mechanic-upgrade", "precedent": [ "eve-online", "second-life", "ultima-online", "mule" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Keep player-driven markets liquid on small servers with model-driven traders whose behavior is calibrated to human trading and constrained by market rules.", "thesis": "Player-driven economies need large populations; EVE Online works because it has them, and smaller persistent worlds stagnate. Agents calibrated to human trading behavior, and analyzed with mean-field methods for stability, can supply liquidity and price discovery without dominating. Mechanism design keeps them from front-running humans.", "type": "opportunity", "validation_experiment": "Simulate a 200-player market with and without calibrated agents using logged order data from a public game economy; measure spread, depth, volatility, and human profit share. Success: tighter spreads and lower volatility with human profit share unchanged within 5 percent." }, { "_derived": { "prediction_status": "open", "priority": 3.587 }, "applies": [ "differentiable-simulation", "llm-explanation-of-state", "constrained-decoding" ], "assertions": [ "derived.production-chain.technique_count == 0" ], "design_sketch": "A bottleneck lens colors every producer by how much one more unit of it would raise the selected output, and a sentence explains the top constraint ('circuits are limited by copper cable, which is starved because the smelter column shares a belt with iron'). Clicking jumps to the machine.", "falsifiable_signal": { "criteria": "Feature first shipped after 2026-10-24 in a released title or official mod tool with at least 1,000 Steam user reviews; patch notes or a developer blog state that it perturbs or re-runs the player's network to identify the limiting resource, machine, or link.", "prediction": "A factory or city-building game will ship a feature that names the binding constraint in the player's production network by testing the simulation, not by displaying static ratios.", "resolve_by": "2028-06-30" }, "feasibility": 5, "horizon": "now", "id": "bottleneck-explainer", "impact": 3, "involves": [ "production-chain", "conveyor-transport", "storage-capacity", "stats-dashboards", "data-overlays" ], "key_risks": [ "Players who enjoy diagnosis lose the puzzle; must be optional or costly.", "Nonlinear buffers make one-step sensitivity misleading.", "Explanations wrong in edge cases erode trust." ], "name": "Bottleneck explainer for production networks", "novelty": 3, "opportunity_type": "legibility-layer", "precedent": [ "factorio", "satisfactory", "anno-1800", "oxygen-not-included" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes", "2026-10-25-contributor-readiness" ] }, "sources": [], "status": "asserted", "summary": "Explain where and why a production network is starved or blocked by combining sensitivity analysis of the actual simulation with grounded language explanations.", "thesis": "No technique in the dataset is linked to production chains, yet one of the mechanic's documented failure modes is diagnosis: starvation far upstream is hard to see, so players blame the wrong building. Differentiable or finite-difference sensitivity of output to each machine, buffer, and route identifies the binding constraint exactly; a constrained language layer turns that into a short explanation with links to the machines involved. The same tool supports designers checking recipe balance.", "type": "opportunity", "validation_experiment": "Implement sensitivity analysis over a factory prototype's production graph and test whether players fix an injected bottleneck faster with the lens. Success: 50 percent faster fixes with no drop in reported satisfaction among experienced players when the lens is optional." }, { "_derived": { "prediction_status": "open", "priority": 4.0 }, "applies": [ "massively-parallel-simulation", "graph-network-simulators", "differentiable-simulation" ], "design_sketch": "Before enacting a policy the player opens Preview: the game forks 32 to 256 copies from the current tick with stochastic seeds, fast-forwards each one in-game year on spare cores (or a learned surrogate), and shows fan charts for key indicators plus the two most divergent futures as short replays. Preview costs an in-game resource so it remains a decision.", "falsifiable_signal": { "criteria": "Feature first shipped after 2026-10-24 in a title with at least 1,000 Steam user reviews; store page, patch notes, or a developer talk describes multiple forward runs from the live save state, shown to the player. A single deterministic projection does not count.", "prediction": "A city builder, colony sim, or grand strategy game will show players several simulated stochastic futures of their current game state before a decision.", "resolve_by": "2028-06-30" }, "feasibility": 4, "horizon": "1-2y", "id": "branch-preview-forecasts", "impact": 4, "involves": [ "forecasting", "policy-toggles", "data-overlays", "stats-dashboards", "variable-speed" ], "key_risks": [ "Previews reveal hidden randomness and can make play feel solved.", "Simulations with large state may not fork cheaply on consumer hardware.", "Surrogate predictions that disagree with the real run break trust." ], "name": "Branch preview of policy futures", "novelty": 4, "opportunity_type": "new-mechanic", "precedent": [ "frostpunk", "democracy-4", "surviving-mars" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Fork the simulation in the background and show the player likely futures of a decision before committing, using parallel or surrogate simulation.", "thesis": "Forecasting in shipped games is shallow: a runway counter, a weather icon, a predicted approval delta. Accelerated and learned simulators make it cheap to run many short futures of the actual game state. Showing a distribution of outcomes for a pending law, budget, or layout change turns opaque coupling into a visible tradeoff and makes long feedback delays playable. Deterministic, event-sourced simulations can fork from any tick, so the preview runs the real rules rather than a hand-built predictor.", "type": "opportunity", "validation_experiment": "Take a deterministic city or colony prototype, implement state snapshot plus N-way fork, and measure fork-and-simulate cost for one in-game year at mid-game scale. Playtest with preview off and on; measure decision regret (outcome versus best available) and session length. Success: preview of 64 branches under 5 seconds on a mid-range PC and a measurable drop in regret without lower engagement." }, { "_derived": { "prediction_status": "open", "priority": 2.764 }, "applies": [ "simulation-based-inference", "differentiable-simulation", "massively-parallel-simulation" ], "design_sketch": "A contagion sim offers a 'calibrated scenario' mode where parameters come from posterior distributions fitted to historical outbreaks. Outcomes show credible intervals; a fit report explains what data the model matches and where it is unreliable.", "falsifiable_signal": { "criteria": "Released after 2026-10-24; release notes, documentation, or an accompanying paper report the calibration method (Bayesian or simulation-based inference), the data it was fitted to, and at least one goodness-of-fit metric.", "prediction": "A playable serious-game or educational simulation will publish posterior-calibrated parameters with fit metrics as part of its release.", "resolve_by": "2028-12-31" }, "feasibility": 4, "horizon": "1-2y", "id": "calibrated-serious-sims", "impact": 2, "involves": [ "contagion", "traffic-congestion", "supply-demand-pricing", "forecasting", "stats-dashboards" ], "key_risks": [ "Users over-trust calibrated outputs.", "Data availability limits domains.", "Calibration may reduce play value." ], "name": "Calibrated serious simulations", "novelty": 3, "opportunity_type": "non-game-transfer", "precedent": [ "plague-inc", "rebel-inc", "cities-skylines", "capitalism" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Use simulation-based inference and differentiable simulators to calibrate game-grade models of epidemics, traffic, and markets to real data, with published fit metrics.", "thesis": "Plague Inc. and Rebel Inc. are used in classrooms and briefings despite stylized models. Simulation-based inference and differentiable simulation make it practical to fit agent-based and compartment models to observed data and quantify uncertainty. Shipping calibrated parameter sets with published fit metrics lets game interfaces carry real insight while keeping play.", "type": "opportunity", "validation_experiment": "Fit a game-grade SIR network model to a public historical outbreak with simulation-based inference and report posterior predictive coverage. Success: 90 percent interval coverage on held-out weeks." }, { "_derived": { "prediction_status": "open", "priority": 4.423 }, "applies": [ "llm-explanation-of-state", "constrained-decoding" ], "design_sketch": "A query box and a right-click 'why?' on any entity. The answer is three to six sentences, each linked to log events the player can jump to on the timeline, plus a small causal graph. Answers that cannot be grounded say so. Advisors use the same pipeline proactively when a key indicator crosses a threshold.", "falsifiable_signal": { "criteria": "Feature first shipped after 2026-10-24 in a title with at least 10,000 Steam user reviews; release notes or the store page describe free-text causal questions answered from the player's own save state, not from a wiki or general help text.", "prediction": "A widely played simulation or strategy game will let players ask in natural language why something happened in their own game and answer from game state.", "resolve_by": "2027-12-31" }, "feasibility": 4, "horizon": "1-2y", "id": "causal-why-query", "impact": 5, "involves": [ "causal-tooltips", "agent-inspection", "advisors", "history-chronicle", "stats-dashboards", "alerts-feed" ], "key_risks": [ "Ungrounded but fluent answers teach players wrong models of the system.", "Event logs in shipped engines often lack cause links; instrumenting them is real work.", "Explanations can make optimal play too easy, flattening difficulty." ], "name": "Why-query over the simulation log", "novelty": 4, "opportunity_type": "legibility-layer", "precedent": [ "dwarf-fortress", "victoria-3", "rimworld", "cities-skylines-2" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Let players ask why something happened and answer from the simulation's own event log, making deep systemic games legible without simplifying them.", "thesis": "The strongest constraint on simulation depth has been legibility: players abandon systems they cannot diagnose, so designers cap coupling. Overlays and causal tooltips help with current values but not with histories ('why did the hospital close?'). If the simulation records an event log with causes, a language model can retrieve the relevant chain and narrate it with citations to specific events, and a constrained output format can force every claim to point at a logged record. This turns the event log, which many engines already keep for replays and debugging, into a player-facing explanation layer.", "type": "opportunity", "validation_experiment": "Instrument an open-source simulation (a city or colony sim) to log events with parent-cause ids. Give two player groups the same failing save; one gets the why-query. Measure time to correct diagnosis and the share of answer sentences whose cited events actually support them (human audit). Success: 40 percent faster diagnosis with at least 95 percent supported sentences." }, { "_derived": { "prediction_status": "open", "priority": 3.318 }, "applies": [ "llm-game-master", "constrained-decoding", "automated-playtesting" ], "design_sketch": "Designers write 200 event templates with preconditions, effects, and tone notes. At build time or runtime, a generator instantiates variants that reference specific characters, places, and history from the save, emits schema-valid events, and passes them through a validator that rejects anything whose effects violate invariants. Generated events are tagged so players can filter them.", "falsifiable_signal": { "criteria": "Feature first shipped after 2026-10-24 in a title with at least 1,000 Steam user reviews; the Steam AI disclosure or developer documentation states that events are generated during play and must conform to a fixed structure of triggers, options, and effects.", "prediction": "A grand strategy, life sim, or narrative management game will ship runtime-generated events constrained by an event schema.", "resolve_by": "2027-12-31" }, "feasibility": 4, "horizon": "now", "id": "content-elastic-events", "impact": 3, "involves": [ "triggered-event-chains", "villager-requests", "dilemma-cards", "procedural-characters", "storyteller-director" ], "key_risks": [ "Generated text quality below authored text lowers perceived quality.", "Disclosure and player backlash against AI-generated writing.", "Validation catches logic errors but not tone errors." ], "name": "Content-elastic event systems", "novelty": 3, "opportunity_type": "authoring-tool", "precedent": [ "crusader-kings-3", "stardew-valley", "animal-crossing-new-horizons", "reigns", "frostpunk" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Remove the authoring ceiling on event chains, requests, and dilemmas by generating schema-constrained variants from designer templates and validating them against game state.", "thesis": "Event chains, villager requests, and dilemma cards are capped by writing budgets: players exhaust the deck and recognize repeats. Language models can generate variants, but unconstrained text breaks game logic. Templates with typed slots, constrained decoding into the event schema, and automated checks against preconditions let a small team ship a deck that keeps producing valid, state-aware events.", "type": "opportunity", "validation_experiment": "Take 50 authored events from an existing design, generate 500 variants, have blind raters compare authored and generated events on quality, and run a validator on every variant. Success: generated events rated within 0.5 points of authored ones with 99 percent passing validation." }, { "_derived": { "prediction_status": "open", "priority": 3.318 }, "applies": [ "llm-game-master", "constrained-decoding" ], "design_sketch": "Each candidate incident from the authored deck is scored on threat budget plus a narrative score: which characters it involves, whether it resolves or escalates an open thread, and whether it pays off a setup. The chosen incident is logged with the narrative reason, readable in the chronicle.", "falsifiable_signal": { "criteria": "Feature first shipped after 2026-10-24 in a title with at least 1,000 Steam user reviews; developer documentation, patch notes, or a technical talk states that a language model scores or ranks candidate authored events against the current game state at runtime. Generated event text alone does not count.", "prediction": "A colony sim or survival management game will ship an event director that uses a language model to rank authored events against the current story state.", "resolve_by": "2028-06-30" }, "feasibility": 4, "horizon": "now", "id": "drama-aware-storyteller", "impact": 3, "involves": [ "storyteller-director", "triggered-event-chains", "relationship-values", "memories-thoughts", "history-chronicle" ], "key_risks": [ "Players perceive arc-shaped events as rigged.", "Narrative scoring adds latency at incident time.", "Arcs can repeat across runs if scoring is too deterministic." ], "name": "Drama-aware storyteller", "novelty": 3, "opportunity_type": "mechanic-upgrade", "precedent": [ "rimworld", "crusader-kings-3", "frostpunk", "this-war-of-mine" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "A storyteller that models characters' arcs and relationships, not only threat budgets, choosing authored incidents that create meaningful stakes for specific colonists.", "thesis": "RimWorld's storytellers select incidents mostly by threat budget and timing. They create tension but not arcs: they do not know that two colonists are estranged siblings or that the doctor is the colony's only hope. A model that reads relationship and memory state can score candidate incidents by narrative relevance while the incidents themselves stay authored and rules-bound, adding arc awareness without surrendering control to generation.", "type": "opportunity", "validation_experiment": "Add narrative scoring to an incident deck in a colony prototype; run blind playthroughs and ask players to recount their run a week later. Success: players recall more named characters and causal links with the narrative storyteller." }, { "_derived": { "prediction_status": "open", "priority": 2.769 }, "applies": [ "genie-world-models", "diffusion-game-engines", "open-interactive-video-models" ], "design_sketch": "Clicking a memory in a colonist's inspector plays a few seconds of generated footage conditioned on the logged event and the colonist's mood. Rumors render with distortions proportional to how many retellings they passed through. Ruins found in worldgen offer a short generated vision of their past from history-generation records.", "falsifiable_signal": { "criteria": "Feature first shipped after 2026-10-24 in a title with at least 1,000 Steam user reviews or a console release; store page, trailer, or patch notes describe generated visuals of past in-game events specific to the player's run. Pre-rendered cutscenes do not count.", "prediction": "A game will use a generative video or world model to show character memories or in-world histories drawn from the simulation's own event log.", "resolve_by": "2028-12-31" }, "feasibility": 3, "horizon": "1-2y", "id": "dream-replay-layer", "impact": 2, "involves": [ "memories-thoughts", "history-chronicle", "gossip-propagation", "history-generation", "forecasting" ], "key_risks": [ "Generation latency breaks the inspection flow.", "Generated imagery of violent events needs content controls.", "Players may read subjective views as ground truth." ], "name": "Generated dream and replay views", "novelty": 5, "opportunity_type": "new-mechanic", "precedent": [ "dwarf-fortress", "rimworld", "kenshi" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Use generative world models for non-authoritative views such as memories, rumors, and alternate histories, where drift is acceptable and variety is the point.", "thesis": "The weaknesses of generative world models (short consistency horizon, no exact state) disqualify them as engines but suit a different role: showing things that are supposed to be subjective. A colonist's memory, a rumor that changed in the telling, a forecast that might not happen, or the history of a ruin are views where approximation reads as character rather than bug.", "type": "opportunity", "validation_experiment": "Prototype generated memory clips from 50 logged events in a colony sim using an available video model; measure generation time and ask players whether clips increased attachment. Success: under 10 seconds per clip and a majority reporting increased attachment." }, { "_derived": { "prediction_status": "open", "priority": 3.776 }, "applies": [ "constrained-decoding", "llm-explanation-of-state", "massively-parallel-simulation" ], "design_sketch": "Developers use a runtime where every state change, including those decided by a model, is an event with inputs, outputs, and causes. Replays and multiplayer clients consume logged outputs instead of re-calling models. Tools offer fork-from-here, diff-between-branches, and cause tracing.", "falsifiable_signal": { "criteria": "Released after 2026-10-24 by any engine or middleware vendor or as an open-source project; documentation describes recording model outputs into a replayable log, and at least one shipped title with at least 1,000 Steam user reviews uses it.", "prediction": "An engine, middleware vendor, or open-source project will ship an event-log replay and determinism layer designed for games with generative AI components.", "resolve_by": "2028-12-31" }, "feasibility": 4, "horizon": "1-2y", "id": "event-sourced-sim-runtime", "impact": 4, "involves": [ "history-chronicle", "forecasting", "persistent-world", "causal-tooltips", "modding-api" ], "key_risks": [ "Log size grows quickly in large simulations.", "Retrofitting existing engines is costly.", "Privacy and storage of logged player text." ], "name": "Event-sourced simulation runtime", "novelty": 3, "opportunity_type": "infrastructure", "precedent": [ "dwarf-fortress", "factorio", "eve-online" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "An engine layer that records simulation and model calls as an append-only event log, giving replay, forking, explanation, and determinism for games with AI components.", "thesis": "Most AI-component risks in simulation games are determinism risks: model outputs are not reproducible, so replays, lockstep multiplayer, bug reports, and save integrity break. Recording every model call and its output as an event, and treating the log as the source of truth with state as a projection, restores reproducibility: replays read cached outputs, forks branch from any event, and explanations cite events. The same log feeds why-queries, branch previews, and chronicles.", "type": "opportunity", "validation_experiment": "Wrap the model calls of an AI-agent prototype in an event log, then replay 100 sessions and verify byte-identical state reconstruction and fork divergence only after the fork point. Success: 100 percent reproducible replays with log overhead under 20 percent of save size per hour." }, { "_derived": { "prediction_status": "open", "priority": 3.318 }, "applies": [ "generalist-game-agents", "llm-explanation-of-state", "model-based-planning-muzero" ], "design_sketch": "Hand a region, department, or supply chain to a manager. It proposes actions with reasons referencing indicators and the player's past choices; the player can approve, veto, or adjust standing preferences. A weekly report summarizes what changed.", "falsifiable_signal": { "criteria": "Feature first shipped after 2026-10-24 in a title with at least 1,000 Steam user reviews; developer documentation states that delegated subsystems adapt to the specific player's recorded choices. Fixed presets or global difficulty settings do not count.", "prediction": "A grand strategy or management game will ship AI managers that learn delegation policy from the individual player's own past decisions.", "resolve_by": "2028-12-31" }, "feasibility": 4, "horizon": "1-2y", "id": "explained-delegation", "impact": 3, "involves": [ "delegation-to-ai", "advisors", "policy-toggles", "stats-dashboards", "staff-hiring" ], "key_risks": [ "Imitating a novice player perpetuates mistakes.", "Explanations that rationalize rather than explain.", "Late-game play becomes approval clicking." ], "name": "Explained delegation to AI managers", "novelty": 3, "opportunity_type": "mechanic-upgrade", "precedent": [ "stellaris", "football-manager", "anno-1800", "victoria-3" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Subsystem managers that learn the player's preferences from their past decisions and explain each action, so delegation scales play without hiding it.", "thesis": "Delegation exists (Stellaris sectors, Football Manager staff, automated trade routes) but managers are weak and opaque, so players either micromanage or tolerate bad decisions. Imitation of the player's own past choices plus explanation grounded in game state makes delegation trustworthy: the manager does what you would do and tells you why.", "type": "opportunity", "validation_experiment": "Train a manager by imitation on a player's first 10 hours of decisions in a strategy prototype, then measure agreement with the player's choices in held-out situations and player trust ratings. Success: 75 percent agreement and higher trust than the baseline AI manager." }, { "_derived": { "prediction_status": "open", "priority": 2.989 }, "applies": [ "constrained-decoding", "llm-game-master", "automated-rule-generation" ], "design_sketch": "Each day introduces rule changes written as formal predicates and rendered as in-world memos. Applicants, documents, and stories are generated to exercise boundary cases of the current rules. The game knows the correct ruling for every case and scores accordingly.", "falsifiable_signal": { "criteria": "Released after 2026-10-24 with at least 1,000 Steam user reviews; store page or developer statement describes runtime-generated cases whose correct outcome is computed from the game's rules, not authored per case.", "prediction": "An inspection or bureaucracy job sim will generate documents or applicants at runtime and score the player against an answer key derived from formal rules.", "resolve_by": "2027-12-31" }, "feasibility": 5, "horizon": "now", "id": "generated-inspection-shifts", "impact": 2, "involves": [ "rule-checking-inspection", "contracts-orders", "dilemma-cards", "procedural-characters" ], "key_risks": [ "Generated text leaks the answer or contradicts the formal case.", "Repetitive tone across generated applicants." ], "name": "Generated inspection shifts", "novelty": 3, "opportunity_type": "mechanic-upgrade", "precedent": [ "papers-please", "contraband-police" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Rule-checking job sims with generated documents, items, and rule changes each shift, verified against an answer key so infinite variety stays fair.", "thesis": "Papers, Please and its successors are limited by hand-built document sets that players memorize. Constrained generation can produce documents, forgeries, and edge cases from a formal rule set, and because the rules are formal, every generated case has a computable correct answer. That combination gives endless fair variety, which ordinary procedural generation struggles to make interesting and unconstrained generation cannot make fair.", "type": "opportunity", "validation_experiment": "Generate 1,000 inspection cases from a 30-predicate rule set and have a checker and two humans verify the answer key on a sample of 200. Success: 99 percent agreement between generated answers and human rulings." }, { "_derived": { "prediction_status": "open", "priority": 2.764 }, "applies": [ "automated-rule-generation", "quality-diversity-search", "automated-playtesting", "alphaevolve-algorithm-discovery" ], "assertions": [ "derived.production-chain.prevalence_rank <= 10", "derived.tech-tree.prevalence_rank <= 10" ], "design_sketch": "Designers set a theme, resource set, and target depth; the tool proposes recipe graphs with cycle times and alternates, simulates optimal and novice factories, and returns an archive ranked by depth, branching, and degeneracy checks. Designers edit and lock items they like.", "falsifiable_signal": { "criteria": "Released after 2026-10-24; a game with at least 1,000 Steam user reviews, or a mod with at least 10,000 downloads on its official mod portal; the developer or mod author states that the shipped recipe graph was produced or substantially shaped by a generative tool.", "prediction": "A factory or crafting game or overhaul mod will credit algorithmic or model-based generation for its recipe graph or tech tree.", "resolve_by": "2028-12-31" }, "feasibility": 4, "horizon": "1-2y", "id": "generated-recipe-graphs", "impact": 2, "involves": [ "production-chain", "tech-tree", "hand-crafting", "milestone-unlocks" ], "key_risks": [ "Generated graphs feel arbitrary without thematic logic.", "Agents optimize for metrics that miss what makes a recipe satisfying." ], "name": "Generated recipe graphs", "novelty": 3, "opportunity_type": "authoring-tool", "precedent": [ "factorio", "satisfactory", "dyson-sphere-program", "minecraft" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes", "2026-10-25-contributor-readiness" ] }, "sources": [], "status": "asserted", "summary": "Generate solvable, non-degenerate, on-theme production recipe graphs and tech trees, verified by agents, to cut content cost for factory and crafting games.", "thesis": "Recipe graphs and tech trees are among the most prevalent content structures in the catalog, and their design is slow: each new item must be reachable, useful, and not dominated. Automated rule generation and quality-diversity search can propose graph variants, and agent playtests can check reachability, ratio complexity, and whether alternate recipes create real choices. Modding communities could use the same tool for overhaul mods.", "type": "opportunity", "validation_experiment": "Generate 100 recipe graphs for a fixed resource set and have experienced factory players rate a sample against an authored graph blind. Success: top-decile generated graphs rated equal to the authored one." }, { "_derived": { "prediction_status": "open", "priority": 3.059 }, "applies": [ "text-to-3d-assets", "gaussian-splatting", "constrained-decoding" ], "design_sketch": "A generate button in build mode takes a prompt or sketch, produces three candidates in the game's style, auto-fits collision and placement metadata, and assigns gameplay tags (seating, light, beauty value) from a classifier. Players share pieces with provenance labels.", "falsifiable_signal": { "criteria": "Feature first shipped after 2026-10-24 in a life or management sim other than inZOI with at least 5,000 Steam user reviews or a console release; store page or patch notes describe player-facing text-to-3D or image-to-3D object generation inside the game.", "prediction": "A life or management sim other than inZOI will let players generate new placeable 3D objects from text or images.", "resolve_by": "2027-12-31" }, "feasibility": 4, "horizon": "now", "id": "generative-decoration-catalogs", "impact": 3, "involves": [ "decoration-beauty", "free-placement", "content-sharing", "room-recognition" ], "key_risks": [ "Copyright and trademark in generated pieces.", "Player and press backlash against generated assets.", "Gameplay tags misassigned to generated objects." ], "name": "Generative decoration catalogs", "novelty": 2, "opportunity_type": "mechanic-upgrade", "precedent": [ "the-sims-4", "planet-coaster", "inzoi", "minecraft", "tiny-glade" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Let players generate furniture, scenery, and building pieces from text or sketches within a game's style and technical budgets, removing the cap on creative-mode catalogs.", "thesis": "Creative-mode sims (The Sims build mode, Planet Coaster scenery, Townscaper-likes) live on catalog breadth, which is bounded by art budgets and DLC schedules. Text-to-3D asset generation reached production use in 2025, and style-constrained pipelines can enforce polygon budgets, collision, and art direction. Generated pieces then feed decoration scores and sharing.", "type": "opportunity", "validation_experiment": "Generate 200 decor items from player prompts within a game's style guide; have artists rate style match and check technical budgets. Success: 80 percent pass technical budgets and a majority rated on-style." }, { "_derived": { "prediction_status": "open", "priority": 3.178 }, "applies": [ "human-behavior-simulation-models", "simulation-based-inference" ], "design_sketch": "Guest archetypes are generated offline from a human-behavior model conditioned on demographics; each archetype yields compact preference parameters the runtime uses. Players see segment-level reasoning ('families find this ride too intense for the price'). A sandbox mode lets designers test offerings against segments.", "falsifiable_signal": { "criteria": "Released or patched after 2026-10-24 in a title with at least 1,000 Steam user reviews; developer blog, talk, or store text attributes in-game demand or preference parameters to a behavioral, survey-fitted, or language model.", "prediction": "A management or tycoon game will state that its customer or citizen preferences come from a model of human behavior learned from data.", "resolve_by": "2028-12-31" }, "feasibility": 3, "horizon": "1-2y", "id": "human-calibrated-customers", "impact": 3, "involves": [ "customer-value-choice", "crowd-preferences", "product-development", "supply-demand-pricing" ], "key_risks": [ "Models encode demographic stereotypes that are offensive or wrong.", "Calibration to real data may make the game less fun than hand-tuned curves.", "Claims of realism invite scrutiny the model cannot survive." ], "name": "Human-calibrated customers", "novelty": 4, "opportunity_type": "mechanic-upgrade", "precedent": [ "rollercoaster-tycoon", "planet-coaster", "game-dev-tycoon", "big-ambitions" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Replace hand-tuned customer and crowd preference curves with models that predict human choices, so tycoon markets respond like real populations.", "thesis": "Customer behavior in management games is a set of hand-tuned curves: price tolerance, ride intensity preference, satisfaction decay. Research in 2024 and 2025 showed language-model agents and choice models trained on large behavioral datasets predicting individual human responses with useful accuracy. Using such models to generate preference distributions, offline, gives tycoon games demand that behaves like real segments and lets players learn transferable intuitions. The same machinery serves product and policy research outside games.", "type": "opportunity", "validation_experiment": "Generate preference parameters for a theme-park prototype from a behavior model, then compare predicted segment choices with a small real survey (n>=300) on the same priced options. Success: rank correlation above 0.6 between predicted and surveyed choices, and playtesters rate guest behavior as more plausible than the hand-tuned baseline." }, { "_derived": { "prediction_status": "open", "priority": 2.756 }, "applies": [ "llm-game-master", "llm-explanation-of-state", "text-to-3d-assets" ], "design_sketch": "At the end of each in-game year, the chronicle drafts a chapter: key events, characters, turning points, written in a selectable voice and illustrated with generated images grounded in the logged entities. Every sentence links to its source events. Players edit and publish chronicles to a shared library.", "falsifiable_signal": { "criteria": "Feature first shipped after 2026-10-24 in a released title or official expansion with at least 1,000 Steam user reviews; patch notes or store text describe an automatically written run history the player can export or share.", "prediction": "A systemic simulation game will ship an automatic, shareable narrative summary of a run's history generated from its event log.", "resolve_by": "2027-12-31" }, "feasibility": 5, "horizon": "now", "id": "illustrated-chronicles", "impact": 2, "involves": [ "history-chronicle", "content-sharing", "history-generation", "procedural-characters" ], "key_risks": [ "Summaries invent drama the log does not support.", "Generated art quality and disclosure.", "Players who enjoy writing their own stories may feel replaced." ], "name": "Illustrated chronicles", "novelty": 2, "opportunity_type": "mechanic-upgrade", "precedent": [ "dwarf-fortress", "rimworld", "crusader-kings-3", "kenshi" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Turn a run's event log into a readable, illustrated saga that players can share, making emergent stories legible and portable.", "thesis": "Players of systemic games already write their runs up as stories (Boatmurdered, colony diaries). Dwarf Fortress's legends mode stores the raw material but presents it as a database. Summarization over event logs, with grounding to cited events and generated illustrations, can produce shareable chapters automatically, which also markets the game.", "type": "opportunity", "validation_experiment": "Generate chronicles for 30 logged colony runs and have their players rate accuracy and whether they would share them. Success: average accuracy above 4 of 5 and more than half willing to share." }, { "_derived": { "prediction_status": "open", "priority": 3.318 }, "applies": [ "constrained-decoding", "on-device-language-models", "llm-explanation-of-state" ], "design_sketch": "An orders panel accepts sentences; each becomes a standing rule shown as a card with the exact priority changes, conditions, and expiry it compiled into. Rules can be edited in the classic UI. Conflicts between rules are flagged with an explanation.", "falsifiable_signal": { "criteria": "Feature first shipped after 2026-10-24 in a title or official expansion with at least 1,000 Steam user reviews; patch notes, store page, or developer statement describe free-text standing orders translated into the game's existing work or priority settings.", "prediction": "A colony or management sim will let players issue standing natural-language orders that compile into its job and priority systems.", "resolve_by": "2028-06-30" }, "feasibility": 4, "horizon": "now", "id": "intent-level-orders", "impact": 3, "involves": [ "work-priorities", "designation-orders", "daily-schedules", "unit-commands", "natural-language-interaction", "delegation-to-ai" ], "key_risks": [ "Ambiguous intents compile to surprising rules.", "Hides the very systems some players enjoy mastering.", "Conflicting standing rules produce opaque behavior." ], "name": "Intent-level orders", "novelty": 3, "opportunity_type": "new-mechanic", "precedent": [ "rimworld", "dwarf-fortress", "oxygen-not-included", "prison-architect" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Players state goals in plain language and the game compiles them into priorities, designations, and schedules, cutting micromanagement in colony and management sims.", "thesis": "Colony and management sims hit a micromanagement wall: the work-priority grid, stockpile filters, and schedules grow until play becomes clerical. Players think in intents ('keep the clinic staffed through winter', 'stop hauling during raids'). Constrained decoding can translate intents into the game's existing control primitives, and the result stays inspectable because it compiles into the same grids the player already understands.", "type": "opportunity", "validation_experiment": "Collect 200 player-stated intents from colony sim forums, compile them into a prototype's control primitives, and have players judge fidelity; measure clicks per in-game day with and without intents. Success: 80 percent judged faithful and a 30 percent drop in management clicks." }, { "_derived": { "prediction_status": "open", "priority": 3.776 }, "applies": [ "automated-rule-generation", "automated-playtesting", "constrained-decoding", "voyager-skill-libraries" ], "design_sketch": "An in-game mod workshop: the player describes the change, the system proposes a diff to data and scripts, runs a headless test scenario with agents to check for crashes and the intended effect, and shows before and after metrics. The player names and shares the mod; the description, code, and test results travel together.", "falsifiable_signal": { "criteria": "Shipped or announced after 2026-10-24 by the developer of a simulation game with at least 10,000 Steam user reviews and an official modding API; the tool outputs mod code or data from text descriptions.", "prediction": "The developer of a moddable simulation game will ship or officially endorse a tool that builds working mods from natural-language descriptions.", "resolve_by": "2028-06-30" }, "feasibility": 4, "horizon": "1-2y", "id": "language-modding", "impact": 4, "involves": [ "modding-api", "content-sharing", "player-programming" ], "key_risks": [ "Generated mods with subtle bugs corrupt saves.", "Security of executing generated code.", "Flood of low-quality mods drowns curation." ], "name": "Mechanic modding by description", "novelty": 3, "opportunity_type": "authoring-tool", "precedent": [ "rimworld", "minecraft", "garrys-mod", "factorio", "cities-skylines" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Players describe a rule change in plain language; a model writes the mod against the modding API and automated agents test it before it is shared.", "thesis": "Modding is how simulation games stay alive for decades, but scripting mods requires programming skill and engine knowledge. Code-generating models write working code against documented APIs, and automated playtesting agents can smoke-test results. Making 'wolves avoid fire' or 'trains prefer cheaper routes' a sentence rather than a project would enlarge the modding population by orders of magnitude.", "type": "opportunity", "validation_experiment": "Collect 100 mod requests from a modding community, generate mods against a documented API, and measure the share that compile, pass a headless scenario, and are judged by the requester to do what they asked. Success: at least 60 percent fully accepted." }, { "_derived": { "prediction_status": "open", "priority": 3.318 }, "applies": [ "wave-function-collapse", "learned-level-generation", "quality-diversity-search" ], "design_sketch": "The player paints a rough block outline and selects intent (dense market street, quiet residential). The assist proposes building footprints and lanes using a learned prior constrained by WFC rules; the simulation scores walkability, land value, and traffic, and the player accepts or tweaks.", "falsifiable_signal": { "criteria": "Feature first shipped after 2026-10-24 in a title with at least 1,000 Steam user reviews; developer documentation states that a model trained on layout data, as opposed to hand-written procedural rules, generates the layout from player brushes or strokes.", "prediction": "A city builder will ship a machine-learned layout generator that turns player strokes into detailed streets or districts.", "resolve_by": "2028-06-30" }, "feasibility": 4, "horizon": "now", "id": "learned-build-assist", "impact": 3, "involves": [ "procedural-build-assist", "zoning", "free-placement", "adjacency-scoring", "land-value" ], "key_risks": [ "Assist removes the planning players enjoy.", "Learned priors import real-world layout biases.", "Unclear ownership of generated layouts." ], "name": "Learned build assist for cities", "novelty": 3, "opportunity_type": "mechanic-upgrade", "precedent": [ "townscaper", "tiny-glade", "manor-lords", "cities-skylines-2" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Extend Townscaper-style procedural assist to full city builders with learned style priors, so rough strokes become plausible districts the simulation then evaluates.", "thesis": "Townscaper, Tiny Glade, and Manor Lords showed that assisted building (the game completes detail around the player's gesture) is joyful. Constraint solving such as wave function collapse covers local consistency; learned layout generators add district-level plausibility. Feeding the result back into the simulation makes assist a gameplay tool rather than only an aesthetic one.", "type": "opportunity", "validation_experiment": "Train a layout generator on open street-network data and compare player-built districts with and without assist on build time and simulated performance. Success: 50 percent less build time at equal or better simulated metrics, with players preferring assist." }, { "_derived": { "prediction_status": "open", "priority": 3.138 }, "applies": [ "graph-network-simulators", "massively-parallel-simulation", "neural-cellular-automata" ], "design_sketch": "A colony or sandbox world where water, heat, gas, and structural stress are simulated everywhere at coarse resolution by a learned surrogate and exactly within a radius of active play, with consistent hand-off at the boundary. Players can flood a valley or collapse a mountain and see effects propagate realistically across the map.", "falsifiable_signal": { "criteria": "Released after 2026-10-24 with at least 1,000 Steam user reviews or a console release; developer documentation or a technical talk confirms that a trained neural model computes physical material behavior during play, not only visual effects.", "prediction": "A game will use a learned surrogate model to simulate fluid, gas, or destruction at runtime.", "resolve_by": "2029-12-31" }, "feasibility": 2, "horizon": "3-5y", "id": "learned-material-surrogates", "impact": 4, "involves": [ "fluid-gas-simulation", "heat-transfer", "voxel-destruction", "falling-sand-materials", "structural-integrity", "soft-body-deformation" ], "key_risks": [ "Surrogates break conservation laws, producing exploitable infinite water or energy.", "Boundary artifacts between surrogate and exact zones.", "Determinism across GPUs for multiplayer." ], "name": "Learned material surrogates at world scale", "novelty": 4, "opportunity_type": "infrastructure", "precedent": [ "noita", "teardown", "oxygen-not-included", "beamng-drive", "the-powder-toy" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Use learned and GPU-parallel physics surrogates to bring per-particle fluids, gases, and destruction from small arenas to large persistent worlds.", "thesis": "Noita, Teardown, BeamNG, and Oxygen Not Included show players love worlds where every particle or voxel is simulated, but compute caps these simulations to small areas or 2D. Graph-network simulators and neural operators learn fluid, cloth, and deformation dynamics at large speedups, and GPU-parallel simulators make the remaining numerical work cheap. Combining a learned surrogate for the far field with exact simulation near the player could extend material simulation to whole maps.", "type": "opportunity", "validation_experiment": "Train a graph-network surrogate on a 2D falling-sand fluid simulator and measure mass conservation error and speedup over 10,000 steps at four times the training domain size. Success: mass error under 1 percent with a speedup above 10 times." }, { "_derived": { "prediction_status": "open", "priority": 3.318 }, "applies": [ "ml-weather-models", "graph-network-simulators" ], "design_sketch": "A farm sim ships a small emulator per biome that rolls forward daily fields (temperature, precipitation, wind) from a seed and a climate scenario. Crops, fire spread, and hydrology read those fields. Players can choose a historical climate or a warming scenario and see droughts and storms arise at realistic frequencies.", "falsifiable_signal": { "criteria": "Feature first shipped after 2026-10-24 in a title with at least 1,000 Steam user reviews or a console release; documentation or a developer statement describes a machine-learned weather emulator running in the shipped game.", "prediction": "A farming, flight, or city sim will ship weather generated by a learned emulator rather than scripted states or live forecast feeds.", "resolve_by": "2028-12-31" }, "feasibility": 4, "horizon": "1-2y", "id": "learned-weather-layer", "impact": 3, "involves": [ "weather", "seasons", "crop-cultivation", "fire-spread", "hydrology", "planetary-climate" ], "key_risks": [ "Global models are too large for clients; distillation to regional scale may lose coherence.", "Realistic weather can be less fun than authored seasons.", "Licensing of model weights and training data." ], "name": "Learned regional weather", "novelty": 3, "opportunity_type": "mechanic-upgrade", "precedent": [ "microsoft-flight-simulator-2020", "farming-simulator-25", "simearth", "timberborn" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Use machine-learned weather emulators to give farming, flight, and city sims realistic, cheap, regionally coherent weather and climate variability.", "thesis": "Game weather is a Markov chain of states or a replay of real forecasts. ML weather models from 2023 to 2025 produce physically coherent global fields orders of magnitude cheaper than numerical prediction, and ensemble methods sample plausible variability. Distilled regional emulators can drive crop growth, fire risk, flooding, and flight conditions with correlated, believable weather, including climate-shift scenarios that sims could never author.", "type": "opportunity", "validation_experiment": "Distill a public ML weather model's outputs into a sub-100 MB regional emulator for one biome; compare statistics (drought frequency, storm clustering) against reanalysis data and run it inside a farming prototype at under 1 ms per in-game day. Success: matched variability statistics within 20 percent and no measurable frame cost." }, { "_derived": { "prediction_status": "open", "priority": 3.318 }, "applies": [ "generative-agents", "on-device-language-models", "massively-parallel-simulation" ], "assertions": [ "rank.whitespace.lifecycle-aging.production-chain <= 10", "cooccur.lifecycle-aging.production-chain <= 1", "derived.production-chain.prevalence_rank <= 10" ], "design_sketch": "Machines need operators with trained skills; skilled workers age and retire; children of workers attend schools the player builds; housing, food, and morale affect shift output. A bottleneck can be a missing machine or a missing mentor. Generational play replaces the usual tech-tree endgame with succession planning for the workforce.", "falsifiable_signal": { "criteria": "Title released after 2026-10-24; its Steam page describes generational workforce mechanics (aging, training, succession) in a factory or automation game, and it has at least 1,000 Steam user reviews by the resolve date.", "prediction": "A factory or automation game in which workers age, train, and are replaced by a simulated next generation will reach 1,000 Steam user reviews.", "resolve_by": "2029-12-31" }, "feasibility": 4, "horizon": "now", "id": "living-workforce-factories", "impact": 3, "involves": [ "production-chain", "lifecycle-aging", "agent-skills", "education-workforce", "needs-decay", "romance-family" ], "key_risks": [ "Workforce management can bury the logistics puzzle players came for.", "Agent counts at factory scale strain compute.", "Two audiences (automation and life-sim) may each find half the game unwanted." ], "name": "Factories with a living workforce", "novelty": 3, "opportunity_type": "new-genre", "precedent": [ "captain-of-industry", "workers-and-resources", "oxygen-not-included", "dwarf-fortress", "banished" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes", "2026-10-25-contributor-readiness" ] }, "sources": [], "status": "asserted", "summary": "Automation games whose workers age, learn, form families, and retire, so a production network is also a society that must be sustained across generations.", "thesis": "The white-space analysis shows production chains and agent lifecycles are both common yet almost never combined: factory games are run by machines or faceless workers, and games with aging, family-forming agents rarely have deep production. Colony sims and Workers & Resources sit closest. A factory whose throughput depends on a workforce with skills, schooling, households, and turnover couples production chains, among the ten most widely used mechanics in the catalog, to the systems that generate the most attachment. Salience-gated language agents make that workforce individually legible without per-tick cost.", "type": "opportunity", "validation_experiment": "Prototype a small automation map where each machine needs a skilled operator drawn from a simulated population with aging and schooling; compare session length and self-reported attachment with a machines-only control. Success: longer sessions and higher attachment without lower throughput-puzzle ratings." }, { "_derived": { "prediction_status": "open", "priority": 3.282 }, "applies": [ "llm-agent-societies", "generative-agents", "on-device-language-models" ], "design_sketch": "The player picks a map, founding traits, and scarcity levels, then seeds 50 to 500 agents. Interventions are god-scale and rare: a drought, a prophet, a trade caravan. A chronicle tool summarizes eras, factions, and turning points; players share worlds and replay them from saved event logs.", "falsifiable_signal": { "criteria": "Title released after 2026-10-24; its Steam page or developer documentation confirms an LLM-driven agent society of at least 50 agents as the core loop, and it has at least 1,000 Steam user reviews by the resolve date.", "prediction": "A game whose core loop is observing and nudging a society of at least 50 language-model agents will reach 1,000 Steam user reviews.", "resolve_by": "2029-12-31" }, "feasibility": 2, "horizon": "3-5y", "id": "llm-society-terrarium", "impact": 4, "involves": [ "persistent-world", "gossip-propagation", "history-chronicle", "player-organizations", "god-powers", "faction-diplomacy" ], "key_risks": [ "Inference cost for hundreds of agents over long runs.", "Agent societies may converge to bland or repetitive behavior without engineered diversity.", "Emergent content safety in open-ended social simulation." ], "name": "Language-agent society terrarium", "novelty": 5, "opportunity_type": "new-genre", "precedent": [ "project-sid", "generative-agents-smallville", "dwarf-fortress", "game-of-life", "black-and-white" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "An observer-gardener genre where players seed and nudge a persistent society of language-model agents and watch institutions, cultures, and histories form.", "thesis": "Generative Agents (2023) and Project Sid (2024) showed that populations of language-model agents produce recognizable social structure: parties, roles, rules, shared beliefs. These are research demos, not games, but the experience they produce resembles the Game of Life and Dwarf Fortress legends: watching a world write its own history. With cheaper inference and chronicle tooling, a terrarium genre becomes viable where the player sets initial conditions and intervenes sparingly, and the product is the history.", "type": "opportunity", "validation_experiment": "Run 100 agents for 30 simulated days on a small local model with a chronicle summarizer; have readers rate 20 generated histories against Dwarf Fortress legends excerpts for interest and coherence. Success: comparable interest ratings at under 1 dollar of inference per simulated day." }, { "_derived": { "prediction_status": "open", "priority": 3.482 }, "applies": [ "on-device-language-models", "constrained-decoding" ], "design_sketch": "The game ships a quantized model for dialogue, event generation, and explanation, with a schema-constrained interface. Higher-quality cloud calls are optional and clearly metered. Mods can swap models.", "falsifiable_signal": { "criteria": "Count titles whose local-model feature first shipped after 2026-10-24; store pages or documentation confirm local inference for runtime features, and each title has more than 1,000 Steam user reviews by the resolve date.", "prediction": "At least five games with over 1,000 Steam user reviews will use an on-device language model at runtime without requiring an internet connection.", "resolve_by": "2027-12-31" }, "feasibility": 4, "horizon": "now", "id": "local-first-ai-sims", "impact": 4, "involves": [ "natural-language-interaction", "generative-content", "procedural-characters" ], "key_risks": [ "Hardware minimums shrink the audience.", "Model quality gap versus cloud models.", "Download size and VRAM competition with rendering." ], "name": "Local-first AI simulation games", "novelty": 2, "opportunity_type": "business-model", "precedent": [ "inzoi", "whispers-from-the-star", "suck-up" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Ship small language and generative models inside the game so AI features run offline with no per-hour inference bill, changing the economics of AI-native simulation.", "thesis": "Cloud inference makes AI features a variable cost per player-hour, which conflicts with the one-time-purchase model of most simulation games and their very long play sessions. Small models running on consumer GPUs and NPUs reached usable quality for constrained tasks in 2025 (inZOI's on-device characters are an example). Local-first design bounds cost, works offline, protects privacy, and survives the vendor's model deprecations.", "type": "opportunity", "validation_experiment": "Benchmark a 1 to 4 billion parameter quantized model on the median Steam hardware survey GPU for schema-constrained NPC dialogue, measuring latency, VRAM, and frame-time impact. Success: under 500 ms first token with less than 10 percent frame-time impact." }, { "_derived": { "prediction_status": "open", "priority": 3.178 }, "applies": [ "mean-field-games", "massively-parallel-simulation" ], "design_sketch": "Traffic and pedestrian flows are fields updated each tick from a mean-field equilibrium solver on the GPU; visible cars and citizens are samples drawn from those fields, so congestion is consistent at scale while inspection still shows individuals with plausible trips.", "falsifiable_signal": { "criteria": "Released or patched after 2026-10-24 in a title with at least 1,000 Steam user reviews; developer documentation names a mean-field, continuum, or equilibrium-field method used at runtime for the main traffic or population model.", "prediction": "A city builder will document a mean-field or continuum crowd model as its primary traffic or population simulation.", "resolve_by": "2029-12-31" }, "feasibility": 3, "horizon": "3-5y", "id": "mean-field-crowds", "impact": 3, "involves": [ "traffic-congestion", "agent-pathfinding", "crowd-preferences", "public-transit", "service-queues" ], "key_risks": [ "Players notice when sampled individuals behave inconsistently with fields.", "Equilibrium solvers can be slow to converge under rapid change.", "Engineering cost to retrofit existing engines." ], "name": "Mean-field crowds and traffic", "novelty": 4, "opportunity_type": "infrastructure", "precedent": [ "cities-skylines-2", "simcity-2013", "cities-skylines", "planet-coaster" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Simulate hundreds of thousands of agents with equilibrium-consistent route and choice behavior by combining mean-field approximations with sampled individuals.", "thesis": "City builders face a ceiling where individually simulated agents cap population and aggregate models feel lifeless. Mean-field game theory describes how large populations of optimizing agents reach equilibrium, and GPU-parallel simulation makes iterating those fields cheap. A hybrid where the mean field sets congestion and choice probabilities and a sample of visible individuals follows them gives large populations that behave coherently.", "type": "opportunity", "validation_experiment": "Implement a mean-field traffic solver on a 100,000-trip city network and compare travel-time distributions with a full agent-based baseline. Success: within 10 percent on mean and 90th percentile travel times at a tenth of the compute." }, { "_derived": { "prediction_status": "open", "priority": 3.282 }, "applies": [ "alphaevolve-algorithm-discovery", "automated-rule-generation", "quality-diversity-search", "automated-playtesting" ], "design_sketch": "A designer selects a base simulation and a family. The system mutates rule code, runs agent populations, scores emergent variety, strategic depth, and legibility proxies, and returns an archive of diverse candidates with replays and plain-language descriptions, deduplicated against the existing taxonomy.", "falsifiable_signal": { "criteria": "Title released after 2026-10-24 with at least 1,000 Steam user reviews or a console release; a developer statement or paper attributes the origin of a shipped rule to automated search over rule sets.", "prediction": "A commercial game will credit one of its shipped mechanics to an automated rule-discovery system.", "resolve_by": "2030-12-31" }, "feasibility": 2, "horizon": "3-5y", "id": "mechanic-discovery-search", "impact": 4, "involves": [ "production-chain", "design-performance-rating", "adaptive-threat-scaling", "modding-api" ], "key_risks": [ "Proxies for fun are weak; search optimizes the proxy.", "Discovered mechanics may be illegible to humans.", "Compute cost of agent evaluation." ], "name": "Automated mechanic discovery", "novelty": 5, "opportunity_type": "authoring-tool", "precedent": [ "factorio", "dwarf-fortress", "spacechem" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Search the space of rule programs with language-model-guided evolution and agent playtests to propose new mechanics that are learnable, deep, and unlike existing ones.", "thesis": "Automated game design has produced board-game rules (Ludii and GAVEL) and levels, and LLM-guided evolutionary search (FunSearch, AlphaEvolve) has found new algorithms and constructions. Simulation mechanics are small rule programs over state; a search that proposes rule variants, evaluates them with agent populations for depth and learnability proxies, and keeps a diversity archive could surface mechanics designers have not tried. The taxonomy in this repository is a candidate search space and novelty reference.", "type": "opportunity", "validation_experiment": "Run LLM-guided evolution over rule variants of a small production-chain sim with a quality-diversity archive, then have designers rate the top 20 candidates for novelty and promise. Success: at least three candidates rated novel and worth prototyping by a majority of designers." }, { "_derived": { "prediction_status": "open", "priority": 3.178 }, "applies": [ "constrained-decoding", "llm-agent-societies", "cicero-strategic-dialogue" ], "design_sketch": "The player drafts 'Ban coal plants within 2 km of housing; subsidize heat pumps by 30 percent.' The compiler returns a structured policy card listing the parameters it maps to and anything it could not map. Interest groups (miners, residents, industrialists) post positions computed from projected welfare changes and voiced by a model. Votes and compliance follow simulation state, not rhetoric.", "falsifiable_signal": { "criteria": "Feature first shipped after 2026-10-24 (early access counts) in a title with at least 1,000 Steam user reviews; documentation shows free-text law input mapped to changes in simulation rules or parameters.", "prediction": "A political or city-building game will let players enact free-text laws that change simulation parameters.", "resolve_by": "2028-12-31" }, "feasibility": 3, "horizon": "1-2y", "id": "natural-language-legislation", "impact": 3, "involves": [ "policy-toggles", "interest-groups", "player-legislation", "natural-language-interaction", "elections", "stability-unrest" ], "key_risks": [ "Players write laws the simulation cannot represent; the unmapped remainder must be visible.", "Group rhetoric can diverge from the computed welfare that actually drives votes.", "Politically sensitive content needs moderation." ], "name": "Natural-language legislation", "novelty": 4, "opportunity_type": "new-mechanic", "precedent": [ "democracy-4", "suzerain", "eco", "a-tale-in-the-desert", "victoria-3" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Players write laws in plain language that are compiled into typed policy parameters, while simulated interest groups argue, lobby, and comply.", "thesis": "Policy mechanics have been menus: a fixed list of toggles chosen by designers. Constrained decoding can compile free-text proposals into a typed policy object over the simulation's actual parameters, and language-model interest groups can respond in character with positions grounded in how the policy affects their simulated welfare. Player-written law moves governance games from choosing among authored options toward authoring rules, which A Tale in the Desert and Eco explored with players as legislators and humans as the parsers.", "type": "opportunity", "validation_experiment": "Build a policy compiler over a 40-parameter city model. Collect 300 player-written laws, have the compiler map them, and have two humans judge mapping fidelity. Success: 85 percent of laws mapped faithfully or with clearly flagged gaps, and playtesters prefer authoring to menus in a within-subjects test." }, { "_derived": { "prediction_status": "open", "priority": 3.0 }, "applies": [ "cicero-strategic-dialogue", "imperfect-information-solvers", "constrained-decoding" ], "design_sketch": "AI rulers send and answer proposals in text; each message is backed by an intent object (offer, demand, threat, promise) that the planner evaluates. Broken promises create named opinion modifiers visible to every faction. Players can still use classic proposal menus; text adds nuance such as conditional offers and side deals.", "falsifiable_signal": { "criteria": "Feature first shipped after 2026-10-24 in a grand strategy or 4X title with at least 5,000 Steam user reviews; developer statement or documentation describes language-based AI diplomacy whose offers are chosen by a planner or game-state evaluation.", "prediction": "A grand strategy or 4X game will ship AI opponents that negotiate in free-text or generated-language diplomacy backed by a planner.", "resolve_by": "2029-06-30" }, "feasibility": 3, "horizon": "1-2y", "id": "negotiating-faction-ai", "impact": 3, "involves": [ "faction-diplomacy", "opinion-modifiers", "reputation", "causal-tooltips" ], "key_risks": [ "Planner cost scales badly with many factions.", "Players exploit language models with persuasion that the planner should ignore.", "Long campaigns exceed the consistency horizon of dialogue memory." ], "name": "Negotiating faction AI", "novelty": 3, "opportunity_type": "mechanic-upgrade", "precedent": [ "crusader-kings-3", "stellaris", "civilization", "europa-universalis-4", "balance-of-power" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Grand strategy rivals that negotiate in language and plan around their promises, combining strategic search with dialogue.", "thesis": "Diplomacy in grand strategy is reduced to acceptance scores on fixed proposals. Cicero showed in 2022 that grounding dialogue in a planner that predicts others' moves lets an agent negotiate and keep useful commitments in a seven-player game. Applying the pattern to faction diplomacy makes alliances, betrayals, and bargains first-class play, and opinion modifiers give the model a structured memory of who kept their word.", "type": "opportunity", "validation_experiment": "In a five-faction diplomacy sandbox, compare menu-only AI with planner-plus-dialogue AI on player-rated plausibility, commitment-keeping rate, and win rates against scripted baselines. Success: higher plausibility and a commitment-keeping rate above 80 percent without exploitable persuasion loops in red-team play." }, { "_derived": { "prediction_status": "open", "priority": 2.5 }, "applies": [ "physics-based-character-control", "massively-parallel-simulation" ], "design_sketch": "Agent animation is driven by a learned controller conditioned on body-model state, load, and mood. The inspector shows the same state the motion expresses. Crowds use cheaper approximations at distance.", "falsifiable_signal": { "criteria": "Feature first shipped after 2026-10-24 in a title with at least 1,000 Steam user reviews; developer documentation names learned motion control (a trained policy driving a physics character) for population agents, not only motion-matched animation.", "prediction": "A management or colony sim will ship learned physics-based motion for its simulated population.", "resolve_by": "2029-06-30" }, "feasibility": 3, "horizon": "1-2y", "id": "neural-npc-motion", "impact": 2, "involves": [ "injury-body-model", "crowd-preferences", "hauling-jobs", "agent-inspection", "rigid-body-physics" ], "key_risks": [ "Runtime cost at crowd scale.", "Uncanny motion worse than stylized animation.", "Physics instability in edge cases." ], "name": "Learned motion for simulated crowds", "novelty": 3, "opportunity_type": "mechanic-upgrade", "precedent": [ "rimworld", "dwarf-fortress", "planet-zoo", "the-sims-4" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Physics-based learned character controllers let simulated agents move and react physically, so crowds, injuries, and labor look like what the simulation says is happening.", "thesis": "Simulation games describe rich internal states (injured, exhausted, carrying heavy loads) but show them with a handful of canned animations, which undermines legibility and attachment. Learned physics-based controllers can produce motion conditioned on state at scale, so an agent with a leg wound limps believably and a tired hauler slows and stumbles.", "type": "opportunity", "validation_experiment": "Drive 200 agents with a learned controller conditioned on three state variables; measure frame cost and whether observers can identify agent state from motion alone. Success: 80 percent identification at under 2 ms per frame." }, { "_derived": { "prediction_status": "open", "priority": 2.648 }, "applies": [ "continuous-alife-and-asal", "neural-cellular-automata", "quality-diversity-search", "massively-parallel-simulation" ], "design_sketch": "The player tunes environment parameters and selection pressures for a continuous artificial-life world. An automated discovery system scans the running world for novel forms and behaviors and bookmarks them; the player curates, names, and transplants organisms between worlds and shares discoveries.", "falsifiable_signal": { "criteria": "Title released after 2026-10-24; its Steam page or developer documentation describes a Lenia-family or neural cellular automaton substrate, and it has at least 1,000 Steam user reviews by the resolve date.", "prediction": "A game built on a continuous artificial-life substrate (Lenia family or neural cellular automata) will reach 1,000 Steam user reviews.", "resolve_by": "2029-12-31" }, "feasibility": 3, "horizon": "1-2y", "id": "open-ended-evolution-sandbox", "impact": 2, "involves": [ "natural-selection", "trait-inheritance", "food-web", "habitat-suitability", "content-sharing" ], "key_risks": [ "Open-ended substrates often stagnate or explode.", "Players may not read abstract organisms as creatures.", "GPU requirements." ], "name": "Open-ended evolution sandbox", "novelty": 4, "opportunity_type": "new-genre", "precedent": [ "species-alre", "spore", "equilinox", "game-of-life", "simlife" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Evolution games built on continuous artificial-life substrates and automated discovery, where morphology and behavior evolve without designer-authored part catalogs.", "thesis": "Evolution games either use authored parts (Spore's editor) or simple genomes (Species, The Bibites). Research on Lenia, Flow-Lenia, neural cellular automata, and foundation-model-guided search for artificial life (ASAL, 2024) produced substrates where lifelike, open-ended forms emerge and tools that find interesting ones automatically. A game can use those tools as a camera operator over an evolving world, surfacing novelty to the player.", "type": "opportunity", "validation_experiment": "Run a Flow-Lenia world on a consumer GPU with an automated novelty detector for 8 hours and have players rate bookmarked discoveries against random snapshots. Success: bookmarked organisms rated more interesting in at least 75 percent of pairwise comparisons." }, { "_derived": { "prediction_status": "open", "priority": 3.059 }, "applies": [ "automated-playtesting", "massively-parallel-simulation", "quality-diversity-search", "differentiable-simulation" ], "design_sketch": "A headless build of the simulation runs on a GPU cluster. A quality-diversity search maintains an archive of strategies across behavior descriptors (expansion speed, military share, trade dependence); a tuner adjusts parameters to keep win rates and resource curves inside designer-set bands; reports flag exploits with replays.", "falsifiable_signal": { "criteria": "Talks, technical blogs, or papers published after 2026-10-24 by staff of two different studios, each naming a commercial title with at least 1,000 Steam user reviews and describing automated agent play at scale used for economy balance.", "prediction": "At least two studios will publicly describe using large-scale agent search to balance a simulation or strategy game's economy before release.", "resolve_by": "2027-12-31" }, "feasibility": 4, "horizon": "now", "id": "parallel-autobalance", "impact": 3, "involves": [ "production-chain", "supply-demand-pricing", "tech-tree", "adaptive-threat-scaling", "rival-firms" ], "key_risks": [ "Bots find exploits humans never would and miss ones humans find.", "Headless builds of legacy engines are expensive to make.", "Over-balancing removes interesting asymmetries." ], "name": "Parallel autobalancing of economies", "novelty": 2, "opportunity_type": "authoring-tool", "precedent": [ "factorio", "offworld-trading-company", "anno-1800", "rimworld" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Run thousands of scripted, learned, and diversity-seeking agents through a game's economy at accelerated speed to find degenerate strategies and tune parameters before players do.", "thesis": "Production chains, markets, threat scaling, and tech trees have many interacting parameters and are tuned by small QA teams and patches after launch. GPU-parallel simulation and quality-diversity search make it practical to run very large populations of strategies and to search parameter space for regions with no dominant strategy. This shifts balance work from reactive patching to pre-launch search.", "type": "opportunity", "validation_experiment": "Run a MAP-Elites search over strategies in an open-source economy game for one GPU-day and compare discovered exploits with the game's historical patch notes. Success: at least half of historically patched economic exploits rediscovered." }, { "_derived": { "prediction_status": "open", "priority": 3.0 }, "applies": [ "simulation-based-inference", "human-behavior-simulation-models", "llm-agent-societies" ], "design_sketch": "A neighborhood is loaded from open data. Participants zone, budget, and route transit; overlays show calibrated effects; simulated residents from different groups react, each tied to the data that grounds them. Sessions export a report of options and tradeoffs.", "falsifiable_signal": { "criteria": "Public planning document or press release dated after 2026-10-24 that names the tool, states that it simulates resident responses, and records its use in an official consultation.", "prediction": "A city government will use a game-style calibrated simulation with simulated resident feedback in a formal planning consultation.", "resolve_by": "2029-12-31" }, "feasibility": 3, "horizon": "1-2y", "id": "participatory-planning-sandboxes", "impact": 3, "involves": [ "zoning", "traffic-congestion", "citizen-satisfaction", "data-overlays", "forecasting", "budget-treasury" ], "key_risks": [ "Simulated residents substitute for real community input.", "False precision of calibrated models.", "Political misuse." ], "name": "Participatory planning sandboxes", "novelty": 3, "opportunity_type": "non-game-transfer", "precedent": [ "cities-skylines", "simcity-1989", "democracy-4" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "City-builder interfaces over calibrated urban models with simulated resident voices, used by communities and planners to explore tradeoffs together.", "thesis": "Planners already use city builders to communicate, but the games' models are not calibrated to real cities and residents are absent. Simulation-based inference can calibrate traffic, housing, and service models to local data; language-model residents grounded in survey data can voice how proposals affect them. The city builder becomes a deliberation interface, with game mechanics (overlays, budgets, forecasts) as the legibility layer.", "type": "opportunity", "validation_experiment": "Calibrate a traffic and housing model for one district, run three participatory sessions, and measure whether participants' understanding of tradeoffs (pre and post quiz) improves relative to a slide presentation. Success: significant improvement with participants rating the tool as fair." }, { "_derived": { "prediction_status": "open", "priority": 3.514 }, "applies": [ "llm-game-master", "constrained-decoding" ], "assertions": [ "rank.whitespace.free-placement.triggered-event-chains <= 5", "cooccur.free-placement.triggered-event-chains <= 1", "derived.triggered-event-chains.prevalence_share >= 0.08" ], "design_sketch": "When the player places or removes a building, the event system evaluates templates against neighbors, inhabitants, and history; a few fire as short vignettes with choices whose effects are typed (happiness, faction, resource). The chronicle records them so a district accumulates a history.", "falsifiable_signal": { "criteria": "Title or expansion released after 2026-10-24 with at least 1,000 Steam user reviews; store page or developer material lists events triggered by player placement decisions that affect specific named inhabitants.", "prediction": "A city builder will advertise placement-triggered narrative events tied to named inhabitants as a core feature.", "resolve_by": "2028-12-31" }, "feasibility": 4, "horizon": "now", "id": "placement-that-tells-stories", "impact": 3, "involves": [ "free-placement", "triggered-event-chains", "adjacency-scoring", "villager-requests", "history-chronicle" ], "key_risks": [ "Event frequency that interrupts building flow.", "Players gaming layout for favorable events.", "Generated vignettes repeat in tone." ], "name": "Placement that tells stories", "novelty": 4, "opportunity_type": "new-mechanic", "precedent": [ "frostpunk", "manor-lords", "islanders", "crusader-kings-3" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes", "2026-10-25-review-round" ] }, "sources": [], "status": "asserted", "summary": "City builders where where and what the player builds triggers specific, state-aware story events about the inhabitants of those buildings.", "thesis": "Building placement and triggered event chains are both common but rarely combined: city builders tell almost no stories about what players build, and event-driven games rarely have spatial building. Events keyed to placement, adjacency, and the inhabitants affected (a tavern next to the temple, a mill that floods a farmer's field) give layout a narrative consequence. Constrained generation can instantiate such events from templates at scale while keeping effects inside the rules.", "type": "opportunity", "validation_experiment": "Add 50 placement-triggered event templates to a small city-builder prototype and measure whether players recall and describe their towns in narrative terms more often than a control. Success: a significant increase in narrative descriptions with no drop in building time." }, { "_derived": { "prediction_status": "open", "priority": 2.756 }, "applies": [ "quality-diversity-search", "learned-level-generation", "automated-playtesting" ], "design_sketch": "A scenario browser shows a grid of challenge types by difficulty; each cell holds generated scenarios that agents have solved, with par times. Players rate them; ratings retrain the generator.", "falsifiable_signal": { "criteria": "Content first shipped after 2026-10-24 in a title with at least 1,000 Steam user reviews; developer documentation names quality-diversity or MAP-Elites generation for shipped scenarios or levels.", "prediction": "A simulation or puzzle-sim game will ship a scenario library generated by quality-diversity search.", "resolve_by": "2028-06-30" }, "feasibility": 5, "horizon": "now", "id": "qd-scenario-libraries", "impact": 2, "involves": [ "scenario-objectives", "procedural-terrain", "milestone-unlocks", "comparative-histograms" ], "key_risks": [ "Generated scenarios feel generic.", "Agent-solvable is not the same as fun." ], "name": "Quality-diversity scenario libraries", "novelty": 2, "opportunity_type": "authoring-tool", "precedent": [ "rollercoaster-tycoon", "poly-bridge", "islanders", "opus-magnum" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Generate large libraries of scenarios that cover a grid of difficulty and play style, so scenario-driven sims never run out of distinct challenges.", "thesis": "Scenario modes are expensive to author and players exhaust them. Quality-diversity search can fill a map of scenario descriptors (resource scarcity, threat timing, terrain constraint) with playable, validated scenarios, and learned generators can add theming. Agent playtests confirm each is solvable and estimate difficulty.", "type": "opportunity", "validation_experiment": "Generate 500 scenarios for a bridge or park builder with MAP-Elites over three descriptors; have players compare random generated scenarios with authored ones. Success: generated scenarios rated at least 80 percent as enjoyable as authored ones." }, { "_derived": { "prediction_status": "open", "priority": 3.776 }, "applies": [ "generative-agents", "on-device-language-models", "constrained-decoding" ], "design_sketch": "Colonists run a standard needs-and-utility loop. When an event crosses a salience score (death of a friend, public insult, near-starvation), the agent gets one model call with its trait list, recent memories, and the event, and must return a schema-valid object: a thought with mood value and duration, an optional relationship change, an optional one-line spoken reaction, and an optional new goal. The player sees these as ordinary thoughts in the inspector, now specific to the colonist's history.", "falsifiable_signal": { "criteria": "Count titles released or patched after 2026-10-24 with more than 1,000 Steam user reviews; developer documentation, postmortems, or mod-verified behavior show model calls triggered by discrete events while routine behavior runs on non-model systems.", "prediction": "At least three colony or life sims with over 1,000 Steam user reviews will call a language model for agents only on salient events rather than per decision.", "resolve_by": "2028-12-31" }, "feasibility": 4, "horizon": "now", "id": "salience-gated-llm-agents", "impact": 4, "involves": [ "utility-action-selection", "needs-decay", "memories-thoughts", "social-exchanges", "relationship-values", "personality-traits" ], "key_risks": [ "Generated thoughts that contradict visible state erode trust faster than generic thoughts do.", "Salience thresholds tuned for drama can flood the log in crisis states.", "Local model quality varies by hardware tier, creating uneven experiences." ], "name": "Salience-gated language agents", "novelty": 3, "opportunity_type": "mechanic-upgrade", "precedent": [ "the-sims", "rimworld", "inzoi", "dwarf-fortress" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Keep cheap utility scoring for every tick and call a language model only at salient moments, giving colony and life sims believable minds at a bounded cost.", "thesis": "Rule-based utility agents are cheap and legible but repetitive; fully LLM-driven agents are expressive but cost tokens every decision and drift from game state. Most agent decisions in The Sims or RimWorld are routine (eat, sleep, haul) and need no language model. The memorable moments are rare: a breakup, a grudge, a mental break, a first meeting. A hybrid that reserves model calls for events above a salience threshold, and writes the result back as typed memories and relationship deltas, captures most of the believability at a small fraction of the cost. Small local models made this affordable in 2025; constrained decoding makes the write-back safe.", "type": "opportunity", "validation_experiment": "Build a 20-colonist prototype with a pure-utility baseline and a salience-gated variant capped at one model call per colonist per in-game hour. Run blind playtests (n>=30) rating believability and attachment, log tokens per player-hour, and count state-contradiction errors per hour. Success: believability gain of at least one point on a 7-point scale at under 2,000 tokens per player-hour with contradiction rate below one per hour." }, { "_derived": { "prediction_status": "open", "priority": 2.764 }, "applies": [ "generative-agents", "on-device-language-models" ], "assertions": [ "derived.tech-tree.prevalence_rank <= 5", "derived.relationship-values.prevalence_share >= 0.08", "rank.whitespace.relationship-values.tech-tree <= 10" ], "design_sketch": "Each project needs a lead and collaborators; compatibility and rivalry change speed and the chance of side discoveries; scientists publish, poach, and defect. The tech graph becomes partially stochastic: what is discovered depends on who works together.", "falsifiable_signal": { "criteria": "Title or expansion released after 2026-10-24 with at least 1,000 Steam user reviews; documentation describes research progress that depends on relationships between named researcher characters, not only on their individual stats.", "prediction": "A strategy or management game will ship research performed by a simulated population of named researchers whose relationships affect outcomes.", "resolve_by": "2029-06-30" }, "feasibility": 4, "horizon": "now", "id": "scientists-not-tech-trees", "impact": 2, "involves": [ "tech-tree", "relationship-values", "agent-skills", "personality-traits", "staff-hiring" ], "key_risks": [ "Stochastic research frustrates players planning builds.", "Micromanaging researchers adds busywork." ], "name": "Research by named scientists", "novelty": 3, "opportunity_type": "mechanic-upgrade", "precedent": [ "stellaris", "civilization", "game-dev-tycoon", "dwarf-fortress" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes", "2026-10-25-contributor-readiness" ] }, "sources": [], "status": "asserted", "summary": "Replace the abstract research meter with named researchers whose skills, relationships, and rivalries determine what gets discovered and when.", "thesis": "Tech trees are among the most widely used mechanics in the catalog and among the least simulated: research is a currency spent on a graph. Relationship systems are common too, but the white-space analysis shows the two rarely meet. Research performed by a small population of named scientists with skills, collaborations, and grudges turns progression into a social system and makes breakthroughs into stories, using the same agent stack as colony sims.", "type": "opportunity", "validation_experiment": "Swap the research meter in a strategy prototype for a ten-scientist system and compare player-recalled research events and build-plan satisfaction. Success: more recalled research stories without lower plan satisfaction." }, { "_derived": { "prediction_status": "open", "priority": 3.059 }, "applies": [ "realtime-voice-models", "llm-explanation-of-state", "on-device-language-models" ], "design_sketch": "During a landing, the instructor says 'airspeed 12 knots fast, reduce power' based on telemetry, then debriefs with a timeline of deviations. Lessons adapt to repeated errors. The player can ask questions mid-flight by voice.", "falsifiable_signal": { "criteria": "Official feature first shipped after 2026-10-24 in a flight or racing simulator with at least 5,000 Steam user reviews or a console release; the instructor speaks generated feedback based on the player's live inputs. Third-party apps do not count.", "prediction": "A flight or racing simulator will ship an AI voice instructor that comments on live telemetry.", "resolve_by": "2027-12-31" }, "feasibility": 4, "horizon": "now", "id": "sim-instructor-coach", "impact": 3, "involves": [ "procedural-checklists", "embodied-skill-control", "flight-aerodynamics", "vehicle-dynamics", "diagnose-repair" ], "key_risks": [ "Wrong advice in safety-relevant domains if transferred to real training.", "Voice latency during time-critical maneuvers.", "Instructor chatter annoys experienced players." ], "name": "Simulator instructor coaching", "novelty": 2, "opportunity_type": "mechanic-upgrade", "precedent": [ "microsoft-flight-simulator-2024", "iracing", "dcs-world", "x-plane", "euro-truck-simulator-2" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Flight, driving, and procedure sims with a voice instructor that watches inputs and state, explains mistakes, and adapts lessons in real time.", "thesis": "Vehicle and job simulators teach through checklists and failure. Professional training uses instructors who watch and talk. Realtime voice models, state-grounded explanation, and player modeling make an instructor that comments on what the player is actually doing, with references to the simulator's telemetry, feasible in consumer sims. The same tool serves professional training and lowers the entry barrier for the most intimidating subgenres.", "type": "opportunity", "validation_experiment": "Add a telemetry-grounded voice instructor to a flight sim landing lesson; compare novice landing success after five attempts with and without the instructor. Success: at least 25 percent higher success with no increase in reported frustration." }, { "_derived": { "prediction_status": "open", "priority": 3.308 }, "applies": [ "generalist-game-agents", "open-ended-environment-generation", "voyager-skill-libraries" ], "design_sketch": "A studio ships a headless build with a typed observation and action API, standard scenarios, and leaderboards for agents. Researchers measure long-horizon performance; players watch agent runs as content.", "falsifiable_signal": { "criteria": "Released after 2026-10-24 by the developer or publisher of a title with at least 1,000 Steam user reviews, as an official announcement or repository; community wrappers and academic forks do not count.", "prediction": "A simulation game developer or publisher will officially release a headless research API or benchmark for AI agents.", "resolve_by": "2027-12-31" }, "feasibility": 5, "horizon": "now", "id": "sims-as-agent-benchmarks", "impact": 3, "involves": [ "production-chain", "power-grid", "conveyor-transport", "scenario-objectives", "comparative-histograms" ], "key_risks": [ "Benchmarks saturate quickly.", "Studios may not want agent play to define their game publicly.", "Engine licensing for research use." ], "name": "Simulation games as agent benchmarks", "novelty": 2, "opportunity_type": "non-game-transfer", "precedent": [ "factorio", "minecraft", "screeps", "dwarf-fortress" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Use factory, colony, and economy sims as long-horizon evaluation and training environments for AI agents, with game studios as environment providers.", "thesis": "AI agents now need environments that test planning over thousands of steps, resource management, and recovery from compounding errors. Simulation games are exactly that, with measurable outcomes (throughput, survival, profit). Minecraft became a research environment through MineDojo and Voyager, and factory games have been proposed as agent benchmarks. Games with headless modes and clean observation APIs can become a licensed research category.", "type": "opportunity", "validation_experiment": "Wrap an open-source colony or factory sim with a typed API, define five scenarios with throughput and survival metrics, and evaluate three frontier agents. Success: scores that separate agents and do not saturate within one model generation." }, { "_derived": { "prediction_status": "open", "priority": 3.062 }, "applies": [ "alphaevolve-algorithm-discovery", "constrained-decoding", "automated-rule-generation" ], "design_sketch": "Each puzzle is a machine to control. The player writes a specification and test cases; a model generates implementations, some subtly wrong; the player's score counts wrong implementations caught, test cost, and spec length, plotted on histograms against other players.", "falsifiable_signal": { "criteria": "Title released after 2026-10-24 with at least 1,000 Steam user reviews; its Steam page describes players specifying, testing, or verifying AI-generated code as the main mechanic.", "prediction": "A programming game will make specifying or testing AI-written code its core loop.", "resolve_by": "2028-06-30" }, "feasibility": 4, "horizon": "now", "id": "specification-programming-games", "impact": 2, "involves": [ "player-programming", "comparative-histograms", "design-performance-rating", "scenario-objectives" ], "key_risks": [ "Niche audience.", "Model behavior changes over versions alter puzzle difficulty.", "Players may find being a reviewer less fun than being a builder." ], "name": "Specification programming games", "novelty": 5, "opportunity_type": "new-genre", "precedent": [ "spacechem", "opus-magnum", "screeps", "bitburner", "the-farmer-was-replaced" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Programming puzzles where an AI writes the code and the player writes the specification and tests, scored on how precisely the spec pins down correct behavior.", "thesis": "Zachtronics games and Screeps made writing code the play. Code generation changes what programming skill means: specifying behavior and catching wrong implementations. A genre built around writing specs and adversarial tests, with a model as an unreliable implementer, teaches the skill that now matters and has a natural scoring rule: how many plausible wrong programs your tests reject.", "type": "opportunity", "validation_experiment": "Ship a 10-puzzle web prototype where a model generates candidate implementations with injected faults; measure completion rates and whether score distributions separate skilled from unskilled testers. Success: a stable leaderboard where test-writing skill predicts score." }, { "_derived": { "prediction_status": "open", "priority": 3.282 }, "applies": [ "diffusion-game-engines", "genie-world-models", "neural-rendering-shaders", "gaussian-splatting" ], "design_sketch": "A city builder whose simulation is a tile grid of typed buildings, agents, and fields. A renderer conditioned on the grid, time of day, and a style prompt produces the view; players restyle their city as medieval, solarpunk, or watercolor without new art. Inspection tools always read the symbolic state.", "falsifiable_signal": { "criteria": "Title released after 2026-10-24 with at least 1,000 Steam user reviews or a console release; developer documentation shows that gameplay state is symbolic and that the main viewport is produced frame by frame by a learned generative renderer.", "prediction": "A game will render its primary view with a generative model conditioned on an explicit symbolic game state.", "resolve_by": "2029-12-31" }, "feasibility": 2, "horizon": "3-5y", "id": "symbolic-sim-neural-render", "impact": 4, "involves": [ "decoration-beauty", "procedural-build-assist", "free-placement", "zoning", "data-overlays" ], "key_risks": [ "Rendered frames that misrepresent symbolic state confuse players (a building that looks like a hospital but is not).", "Inference cost per frame on consumer GPUs.", "Temporal flicker and style drift across sessions." ], "name": "Symbolic simulation, neural rendering", "novelty": 5, "opportunity_type": "infrastructure", "precedent": [ "townscaper", "tiny-glade", "cities-skylines", "islanders" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Keep gameplay in a small deterministic symbolic simulation and let a generative world model render it, decoupling visual variety from simulation cost and art budgets.", "thesis": "Generative world models (GameNGen, Genie 3, Muse, Oasis) produce convincing interactive frames but drift, forget, and cannot be trusted as game state. Simulation games need the opposite: exact, persistent, inspectable state. Splitting the stack resolves the conflict: the authoritative simulation stays symbolic (grids, agents, stocks), and a conditioned generative model renders it, so visual variety becomes nearly free while rules, saves, replays, and multiplayer sync remain deterministic.", "type": "opportunity", "validation_experiment": "Condition an open video or image model on a 32x32 symbolic city grid and measure per-tile semantic fidelity (does a rendered tile read as its type to human raters) and frame time. Success: over 95 percent rater agreement with the symbolic type at 10 frames per second on a current consumer GPU." }, { "_derived": { "prediction_status": "open", "priority": 3.178 }, "applies": [ "generalist-game-agents", "voyager-skill-libraries", "on-device-language-models" ], "design_sketch": "The player demonstrates 'set up a smelting column' once while narrating; the companion abstracts it into a named skill with parameters, shows the generated plan, and reuses it elsewhere. Skills can be edited, combined, and shared. The companion asks clarifying questions when a skill fails.", "falsifiable_signal": { "criteria": "Feature first shipped after 2026-10-24 in a title with at least 1,000 Steam user reviews; store page or developer documentation describes companion skills learned from the player that persist and can be reused on new tasks.", "prediction": "An automation or sandbox game will ship an in-game companion that learns reusable skills from player demonstration or instruction.", "resolve_by": "2028-12-31" }, "feasibility": 3, "horizon": "1-2y", "id": "teachable-sim-companion", "impact": 3, "involves": [ "delegation-to-ai", "blueprints", "player-programming", "natural-language-interaction" ], "key_risks": [ "Agents that play the game for you can remove the core pleasure.", "Skill abstraction from one demonstration often over- or under-generalizes.", "Runtime cost of a capable agent on consumer hardware." ], "name": "Teachable sim companion", "novelty": 4, "opportunity_type": "new-mechanic", "precedent": [ "autonauts", "factorio", "minecraft", "satisfactory" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "A co-player agent that learns factory, colony, or farm routines from demonstration and language, and grows a library of reusable skills the player can inspect.", "thesis": "Generalist game agents (SIMA, SIMA 2) follow language instructions across 3D games, and Voyager showed skill libraries of reusable code that grow over play. Automation games already reward teaching: Autonauts made programming bots by demonstration the core loop. A companion that learns skills from the player and stores them as inspectable, shareable routines turns automation from wiring to mentoring.", "type": "opportunity", "validation_experiment": "In a voxel automation prototype, have players teach five routines by demonstration and narration; measure how often the learned skill succeeds in a new location and whether players report more or less enjoyment than blueprints. Success: 70 percent transfer success and preference over blueprints for at least half of players." }, { "_derived": { "prediction_status": "open", "priority": 3.0 }, "applies": [ "model-based-planning-muzero", "on-device-language-models", "generalist-game-agents" ], "design_sketch": "The player raises a creature in a small world. Rewards, punishments, demonstrations, and spoken commands shape its policy. A 'what I learned' panel lists the creature's inferred rules in plain language with examples; the player can correct misunderstandings. Creatures can be traded and bred, inheriting learned dispositions partially.", "falsifiable_signal": { "criteria": "Title released after 2026-10-24 with at least 1,000 Steam user reviews or a console release; developer documentation states that the creature's behavior comes from a deep neural network (more than one hidden layer), a language model, or a deep RL policy whose parameters change in response to that player's feedback. Single-layer learners such as Black & White's perceptrons do not count.", "prediction": "A virtual-pet or god game will ship a creature controlled by a learned neural or reinforcement-learning policy that player feedback updates during play.", "resolve_by": "2029-06-30" }, "feasibility": 3, "horizon": "1-2y", "id": "trainable-creatures-revival", "impact": 3, "involves": [ "creature-learning", "creature-care", "natural-language-interaction", "trait-inheritance", "god-powers" ], "key_risks": [ "Learning that is too slow or too fast reads as fake.", "Reward hacking produces creatures that game the player.", "On-device training cost." ], "name": "Trainable creatures revival", "novelty": 3, "opportunity_type": "mechanic-upgrade", "precedent": [ "creatures", "black-and-white", "seaman", "nintendogs", "tamagotchi" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes" ] }, "sources": [], "status": "asserted", "summary": "Revive the Creatures and Black & White lineage with creatures whose behavior is genuinely learned from player feedback and can explain itself.", "thesis": "Creatures (1996) used neural networks and genetics; Black & White (2001) let a creature learn from slaps and strokes. Both were limited by tiny learning systems that generalized poorly and were opaque. Modern reinforcement learning, preference learning, and small language models allow a creature that learns real policies from feedback, generalizes to new situations, and can say in simple words what it thinks it was rewarded for.", "type": "opportunity", "validation_experiment": "Build a creature with a learned policy trained online from player feedback in a 2D world; measure how many feedback events it takes to learn three behaviors and whether its self-explanations match its actual policy. Success: under 30 feedback events per behavior and 80 percent explanation accuracy." }, { "_derived": {}, "applies": [ "ml-weather-models", "simulation-based-inference" ], "assertions": [ "derived.power-grid.prevalence_share >= 0.08", "derived.weather.prevalence_share >= 0.08", "cooccur.power-grid.weather <= 1" ], "deprecated": { "reason": "Cities: Skylines II already couples solar and wind output to weather with battery storage, so the prediction was true at writing; the remaining idea (realistic learned weather driving systems) is covered by learned-weather-layer.", "superseded_by": "learned-weather-layer" }, "design_sketch": "Solar and wind output follow simulated weather; heating and cooling demand follow temperature; the player builds storage, buys forecasts, and sets demand-response policies. Rare long calm, cold spells test the grid. Prices respond to scarcity if the game has markets.", "falsifiable_signal": { "criteria": "Developer documentation describing weather-dependent renewable output and storage in the grid system.", "prediction": "A commercial city builder or management game will couple renewable generation to simulated weather as a core grid mechanic.", "resolve_by": "2028-12-31" }, "feasibility": 4, "horizon": "1-2y", "id": "weather-coupled-grids", "impact": 2, "involves": [ "power-grid", "weather", "forecasting", "supply-demand-pricing", "storage-capacity" ], "key_risks": [ "Realistic intermittency may frustrate players without good forecasting tools.", "Weather that dominates the economy reduces player agency." ], "name": "Weather-coupled energy grids", "novelty": 3, "opportunity_type": "mechanic-upgrade", "precedent": [ "cities-skylines-2", "anno-1800", "timberborn", "surviving-mars" ], "provenance": { "created": "2026-10-24-synthesis", "reviewed": [ "2026-10-24-review-fixes", "2026-10-25-contributor-readiness" ] }, "sources": [], "status": "asserted", "summary": "Power grids whose renewable supply follows realistic weather, so storage, forecasting, and demand response become the grid puzzle.", "thesis": "Power grids and weather are both common in the catalog but almost never coupled; most games treat generation as constant. Learned weather emulators make realistic, correlated wind, sun, and temperature affordable, and the coupling creates a familiar modern problem: balancing intermittent supply with storage, forecasts, and flexible demand.", "type": "opportunity", "validation_experiment": "Couple a city prototype's grid to a weather emulator and test whether players with forecast tools build mixed storage strategies rather than one dominant plant type. Success: strategy diversity increases and blackout frustration ratings stay below the baseline game's disaster ratings." } ]