Hostile Neural Networks guide for All the Mods 10 0.45

Hostile Neural Networks is a mob farming mod without the mob farm. Instead of building a spawner room, you capture data about an entity, train that data into a model, and then run simulations that spit out the mob's drops. It's a spiritual successor to Deep Mob Learning, and this pack ships version 6.0.0.^[1][2]

The core item is the Data Model, which stores information about a single entity type. You start by crafting a Data Model Framework and using it on the target mob to build the model. Fresh models begin in the Faulty state and need training before they're useful.^[1]

There are two ways to train. A Deep Learner holds a few models at once and trains them passively from the kills you make yourself, which is how you get a model from Faulty up to Basic. Once a model reaches Basic or higher, you can move it into a Simulation Chamber and train it there instead.^[1]

Turning models into loot

The Simulation Chamber runs simulations using energy and a catalyst item, usually a Prediction Matrix. Every simulation produces a generalized prediction, and depending on the model's accuracy it may also produce a Mob Prediction. Higher model tiers mean better accuracy, topping out at 99.5% at the Self Aware tier.^[1]

Mob Predictions aren't drops yet. Feed them into a Loot Fabricator, which uses power to convert each prediction into an actual mob drop chosen from a pre-defined list for that model type. You pick a target output per model type, and the fabricator remembers your choices, so one machine can handle predictions from several different models.^[1]

Where it fits

This is a mid-game automation mod: you need a steady power supply and a source of catalysts before the simulation loop pays off, but once it's running you get renewable mob drops without any spawning, lighting or mob-cap headaches. It's a good fit if you hate building mob farms or you need a specific drop in bulk.^[1]

Unverified The exact list of Data Models available in this pack, the energy costs of each simulation, and any pack-specific recipe changes are not covered by the sources used for this page.

Hostile Neural Networks lets you simulate mobs instead of farming them, turning data about an entity into real drops. The whole mod revolves around one item: the Data Model, which stores information about a mob so you can generate its loot later.^[1]

Step 1: Get your first Data Model

Before you can model anything, you need a Data Model Framework. Craft one, then use it on the mob you want to study and it becomes a Data Model for that entity. Pick something you actually want drops from, because each model only covers its own mob type.^[1]

Every model starts out in the Faulty state, so a fresh model won't do anything useful until you train it.^[1]

Step 2: Train it to Basic with a Deep Learner

The Deep Learner holds a few Data Models at once, and kills you make yourself feed training data into them. Every model has to be dragged from Faulty up to Basic this way before the automated route opens up, so keep the Deep Learner on you while you're out fighting.^[1]

While you're holding the Deep Learner it shows an overlay listing the models inside, and hovering a model and holding shift expands the tooltip with details like how much experience it generates and how many loot table copies it rolls.^[3]

Step 3: Run simulations

Once a model reaches Basic or higher, you can move it into a Simulation Chamber. Simulations need energy, with the cost depending on the model type, plus a catalyst item in the chamber, usually a Prediction Matrix. Make sure you have power ready before you start.^[1]

Every simulation produces a generalized prediction, and it may also produce a Mob Prediction depending on the model's accuracy. Higher model tiers mean better accuracy, topping out at 99.5% at Self Aware, so it pays to keep training a model even after it starts producing.^[1]

Step 4: Turn predictions into loot

Mob Predictions aren't drops yet. Feed them into a Loot Fabricator, which uses power to convert them into actual mob drops chosen from a pre-defined list. You pick a target item per model type, and the fabricator remembers your choice for each type, so a single machine can handle predictions from several different models.^[1]

Converting generalized predictions

The generalized predictions you get from every simulation aren't dead weight either. If you're running overworld models but need nether or end materials, you can convert them up a dimension at a time on a crafting table.^[1][4]

Generalized Nether PredictionMod recipe

Crafting Shaped

Crafting Shaped

Empty slot	Generalized Overworld Prediction	Empty slot
Generalized Overworld Prediction	Netherrack	Generalized Overworld Prediction
Empty slot	Generalized Overworld Prediction	Empty slot

Recipe id: hostilenetworks:living_matter/overworld_to_nether

Overworld to Nether

Generalized Ender PredictionMod recipe

Crafting Shaped

Crafting Shaped

Empty slot	Generalized Nether Prediction	Empty slot
Generalized Nether Prediction	End Stone	Generalized Nether Prediction
Empty slot	Generalized Nether Prediction	Empty slot

Recipe id: hostilenetworks:living_matter/nether_to_ender

Nether to Ender

Unverified The exact in-pack crafting recipes for the Data Model Framework, Deep Learner, Simulation Chamber, Loot Fabricator and Prediction Matrix aren't covered by the sources used for this page. Check JEI in game for the pack's current versions. This site does hold All the Mods 10's own recipes, read from the pack itself: look the item up in the item browser.

Hostile Neural Networks is built around one idea: you train a model of a mob, simulate it, then turn the results into that mob's drops. Here is how each stage works, as far as the sources explain it.^[1]

The Data Model

The Data Model is the core item of the mod. It holds information about one entity, and you use it to generate that entity's drops. To get one, you craft a Data Model Framework and use it on the target entity to build the model. Then you have to train it.^[1]

Models are defined by datapack files, one file per model in the data_models subfolder. That means other mods or datapacks can add more models.^[1]

Training: the Deep Learner

You can train a model with a Deep Learner or with a Simulation Chamber. The Deep Learner stores a few data models, and the kills you make start training them. Every model starts in the Faulty state and has to be trained up to Basic in the Deep Learner.^[1]

The Deep Learner's tooltip lists the models inside it, and it displays something when held while models are within it.^[3]

The mod also has an attunement feature, where you click on an entity with a certain item to attune a model. The tooltip text shows that it can be disabled, so it may not be available on every setup.^[3]

Simulation Chamber

Once a model is Basic or higher, you can run simulations on it in the Simulation Chamber. Every simulation produces a generalized prediction. It may also produce a Mob Prediction, depending on the model's accuracy. A higher model tier means higher accuracy, up to 99.5% at Self Aware.^[1]

Simulations need energy, and the amount depends on the model type. They also need a catalyst item, which is usually a Prediction Matrix.^[1]

Loot Fabricator

Mob Predictions are processed in a Loot Fabricator. It uses power to turn predictions into mob drops. The possible drops come from a pre-defined list of items, and you can pick a target for each type of model.^[1]

The fabricator remembers your choice for each type, so one fabricator can take several types of predictions.^[1]

Moving up a dimension

This pack has crafting recipes that convert the generalized predictions between dimensions. You can turn a Generalized Overworld Prediction into a Generalized Nether Prediction, and a Nether one into a Generalized Ender Prediction. These are the recipes the pack uses:^[4]

Generalized Nether PredictionMod recipe

Crafting Shaped

Crafting Shaped

Empty slot	Generalized Overworld Prediction	Empty slot
Generalized Overworld Prediction	Netherrack	Generalized Overworld Prediction
Empty slot	Generalized Overworld Prediction	Empty slot

Recipe id: hostilenetworks:living_matter/overworld_to_nether

Generalized Ender PredictionMod recipe

Crafting Shaped

Crafting Shaped

Empty slot	Generalized Nether Prediction	Empty slot
Generalized Nether Prediction	End Stone	Generalized Nether Prediction
Empty slot	Generalized Nether Prediction	Empty slot

Recipe id: hostilenetworks:living_matter/nether_to_ender

Unverified The sources don't give exact energy costs for the Simulation Chamber or Loot Fabricator, the accuracy of tiers below Self Aware, or how the machines handle automation (hoppers, pipes or item and energy sides). No config files for this mod were found in the pack either, so check in game before you build an automated setup.

Hostile Neural Networks in this pack has three prediction tiers, one for each dimension: Generalized Overworld Prediction, Generalized Nether Prediction and Generalized Ender Prediction. They form a chain, and you move up it with the pack's own crafting recipes.^[4]

Overworld to Nether

To get from the Overworld tier to the Nether tier, craft Generalized Nether Prediction from the Overworld prediction. This is a shaped crafting recipe, and here is the exact in-pack version:^[4]

Generalized Nether PredictionMod recipe

Crafting Shaped

Crafting Shaped

Empty slot	Generalized Overworld Prediction	Empty slot
Generalized Overworld Prediction	Netherrack	Generalized Overworld Prediction
Empty slot	Generalized Overworld Prediction	Empty slot

Recipe id: hostilenetworks:living_matter/overworld_to_nether

Nether to End

The last step works the same way. You craft Generalized Ender Prediction from the Nether prediction, again with a shaped crafting recipe:^[4]

Generalized Ender PredictionMod recipe

Crafting Shaped

Crafting Shaped

Empty slot	Generalized Nether Prediction	Empty slot
Generalized Nether Prediction	End Stone	Generalized Nether Prediction
Empty slot	Generalized Nether Prediction	Empty slot

Recipe id: hostilenetworks:living_matter/nether_to_ender

Unverified I couldn't confirm anything else about progression from the available sources: which mobs belong to each tier, how much data or living matter each tier needs, and any quest guidance. Check those in-game before relying on them.

The mod's own project page describes one built-in way to connect with other mods: datapacks. Custom Data Models can be added through a datapack, and each Data Model is a single file in the data_models subfolder.^[1]

The base mod includes a Witch Data Model file as an example of the format. The page also says published datapacks for certain mods may be listed in the future, but it doesn't name any, so don't assume a particular mod's mobs are covered.^[1]

Unverified No in-pack recipes or quests that use items from both this mod and another mod were available, so any specific cross-mod connection in this pack is unverified. The same goes for whether the pack ships a datapack that adds Data Models for mobs from other mods.

Same as in the All the Mods 10 8.1 guide

The mod's own project page describes a few numbers and costs you'll deal with most, but it doesn't list any config options. It says running a simulation in the Simulation Chamber costs energy, and the amount depends on the model type. Each simulation also uses up a catalyst item, usually a Prediction Matrix. It also says accuracy goes up as your Data Model reaches higher tiers, topping out at 99.5% at Self Aware.^[1]

Keep in mind that this is how the mod describes itself in general, not a confirmed All the Mods 10 8.1 setting. The pack could use different values.^[1]

Unverified We couldn't check this pack's Hostile Neural Networks config values, so they aren't listed here. That means we can't tell you whether the pack changed any mod defaults, such as simulation energy costs, accuracy per model tier, or Deep Learner and Simulation Chamber settings. We also couldn't confirm which of these are configurable in the first place. To see the exact values, open the mod's config file in your own All the Mods 10 8.1 instance.