Automation as understanding

Automation in Voxamine is not a reward for reaching a tier and not an excuse to replace a process with a recipe. It is an amplifier for a process the player has already operated, measured, and understood by hand. The factory does not erase chemistry; it makes the chemistry visible at a scale where flow, waste, energy, impurity, and control become the next questions.

This is a design chapter. Belts, pipes, multi-block machines, process controllers, and network logistics are planned systems; the current runtime proves the earlier foundation with player-operated vessels, material batches, and normal crafting. The intended design is still binding: planned automation must not be described as live gameplay until its owning systems exist.

The threshold for automation

An operation becomes a candidate for automation when repeating it manually no longer teaches the player anything new, but the real process still has meaningful constraints. The right question is not “can this be automated?” It is “what does automation expose that hand work hides?”

Candidate	Why it earns automation	What must remain real
Crushing and milling	Finer feed increases reaction surface area; repeated hand processing is labor, not new inference.	Feed mass, particle size, dust, drive power, and machine throughput.
Screening and density separation	Repeated sorting makes ore grade and gangue visible at production scale.	Material composition, rejected fraction, overflow, and recovery.
Furnace or vessel heat treatment	A controlled ramp, soak, and cool profile is more precise than watching a fire.	Heat input, vessel limits, gas pressure, reaction progress, and safe shutdown.
Solids transport	A line can move measured material between stages faster and more consistently than the player.	Each physical unit’s mass, composition, origin, capacity, and blockage.
Fluid and gas routing	Piping makes pressure, containment, and environmental consequence operational problems.	Volume, composition, pressure, leakage, relief, and compatible material.
Quality routing	A known assay can direct high-grade feed or off-spec product to the right destination.	The measurement’s uncertainty and the actual grade/purity threshold.

By contrast, the first run of an unfamiliar mineral, a new reaction, or an uncharacterised apparatus remains manual. The player needs observation before delegation: heat, colour, pressure, gas, mass, and instrument readings come before a controller is asked to reproduce a procedure.

Automation follows a measured loop

flowchart LR
  EXPERIMENT["Manual experiment\nobserve the process"]
  MODEL["Notebook + assay\nstate a hypothesis"]
  LINE["Small production line\nmove real matter"]
  MEASURE["Sensors + cues\ncompare expected flow"]
  CONTROL["Schedule + controller\nact on readings"]
  RESULT["Product, waste, heat, gas\naccount every output"]
  REVISE["Diagnosis\nchange one variable"]

  EXPERIMENT --> MODEL --> LINE --> MEASURE --> CONTROL --> RESULT --> REVISE --> EXPERIMENT

The loop is deliberately circular. A balanced reaction supplies theoretical material demand, not guaranteed output. Real throughput includes ore grade, incomplete conversion, competing reactions, heat transfer, transport capacity, and the limits of the apparatus. The player uses an automated line to see the difference between the stoichiometric prediction and the measured production rate, then has a reason to inspect the bottleneck.

An automated system must therefore report enough to be falsifiable. “Copper line stopped” is not a diagnosis. “Roaster is starved,” “waste outlet is full,” “drive power is limited,” “pressure interlock opened,” and “input is below the requested grade” are meaningful because they correspond to named physical states or measured conditions.

What the player automates first

The intended ladder begins with low-energy, legible mechanics:

1. Gravity chutes move solids downhill. They are free in energy terms but constrained by elevation and routing.
2. Mechanically driven belts move discrete containers, bars, sacks, and batches. Their throughput is derived from unit mass and speed rather than authored as an arbitrary items-per-minute number.
3. Shafts, gearing, and bellows transfer muscle, water, or wind power to a machine whose physical transform has a reason to exist.
4. Sensors and threshold control let a line react to a measured temperature, pressure, level, pH, gas condition, current, or voltage.
5. Programmed profiles and PID control hold a process through a ramp, soak, cool, or unstable operating region that a human cannot regulate reliably.
6. Electric network logistics and drones arrive only once motors, batteries, and current make them physically possible. They solve awkward, distant, or hazardous work, not bulk transport that a belt should handle better.

Automation becomes mandatory only when human response time or endurance is genuinely insufficient. Industrial electrolysis and millisecond-scale plasma control require automation because reality requires it, not because the game wants another gate.

A factory never turns matter into counters

The project’s material rule applies more strongly under automation, not less. Belts carry physical units whose payloads retain mass and composition. A pipe carries a fluid or gas with composition and pressure. A machine buffer contains a bounded physical inventory. A controller may select a destination based on grade or purity, but it may not silently convert an off-spec batch into a standard item stack.

Every boundary has an answer to five questions:

1. What exact payload entered: mass, composition, temperature, and provenance where applicable?
2. Which owner accepted it, and did the transfer commit once?
3. What capacity, direction, and compatibility rule governs the connection?
4. Where does material go when the destination is full, power is lost, or the connection is removed?
5. Which named reservoir, buffer, atmosphere, or physical drop accounts for every output?

This is why a belt jam is content, not an error screen. It reveals that a downstream buffer is full, a waste route is absent, a quality rule has rejected material, or a drive is underpowered. The line stopped for a reason the player can investigate.

Controls are procedures, not magic toggles

Real processes are schedules. A useful automation interface lets the player specify a temperature ramp, soak target, soak duration, controlled cool, feed condition, and safety limits. The procedure can be saved and reused, but it begins as untested. A schedule says what the player intends an apparatus to do; it does not certify that the vessel can survive, the reaction will proceed, or the outputs are safe.

Sensors publish simulated readings. Threshold logic is the first controller: start cooling above a temperature, close a feed at a level, or stop a heater at pressure. Later, proportional and PID controllers make tuning a real part of play. Poor gains should produce an observable oscillation or slow response, and the player should be able to compare that behavior against the target and adjust one control variable at a time.

Safeties are automation too, but they are not decorative insurance. Relief valves, rupture discs, high-temperature cutoffs, gas-detection shutdowns, and emergency stops must act on particular physical paths. An emergency stop cannot delete stored heat, pressure, charge, or material. It can only command the actuators it actually controls, leaving the remaining state available for observation and recovery.

Commission a line before trusting it

The first production run is a structured experiment:

1. Feed a small, measured charge and record the expected material and energy balance.
2. Make every output path visible, including product, residue, waste, heat, and gas.
3. Start with manual or threshold control so the player can see each cause and effect.
4. Compare measured flow against the theoretical ratio and assay the outputs.
5. Add one automation feature at a time: transport, feed, heating, routing, then closed-loop control.
6. Test failure deliberately: fill an outlet, interrupt power, starve an input, disconnect a loaded route, and restart under the declared reset policy.

A line is not proven because it ran unattended in ideal conditions. It is proven when it handles expected starvation, blockage, interruption, and restart conservatively, legibly, and without duplicating or deleting matter.

Useful workflows, from first line to remote work

The following are planned workflow patterns, not a list of live machines. Each one earns automation because it replaces repeated handling with a physical system the player can inspect, tune, and recover. They are useful precisely because they retain the reasons a line succeeds or fails.

1. Copper from ore and wood

This is the first complete industrial workflow and the C4 acceptance target: convert ore and wood into copper hands-off for a sustained run. It begins only after the player has manually recognised, prepared, roasted, and reduced the relevant materials.

Stage	Planned responsibility	Useful observation	Failure the player can recover from
Feed preparation	A crusher/mill, screen, and jig reduce particle size and separate a grade-appropriate feed.	Feed rate, particle-size class, assay, and rejected gangue.	A full reject bin or low-grade feed starves the next stage.
Fuel preparation	Chutes or belts bring wood/charcoal to a bounded buffer.	Fuel mass, buffer level, and mechanical drive load.	A depleted buffer limits heat or forced air without changing the ore.
Thermal process	A driven bellows and controlled apparatus hold the selected roasting/reduction conditions.	Temperature, pressure, exhaust path, and the expected versus measured conversion.	Lost drive power or an interlock stops an actuator while stored heat and charge remain real.
Product handling	The line separates copper-bearing product, slag/residue, and off-gas routes.	Mass balance, product assay, waste capacity, and visible exhaust treatment.	A blocked output stalls upstream work or follows its declared spill/relief path.

The useful lesson is not “place four machines to obtain copper.” It is that a copper rate is constrained by the slowest physical stage. Improving feed grade, milling finer material, adding transport capacity, or supplying more drive only helps when the measurement says that stage is the bottleneck.

2. Heat-treatment and vessel procedures

Many useful operations are not a continuous material line but a repeatable procedure: calcining, drying, annealing, distillation, or a controlled reaction charge. Here automation starts as a recorded temperature and actuator schedule rather than as unattended bulk production.

1. The player manually establishes a safe charge size and observes the apparatus response.
2. They record a ramp, hold, and controlled-cool procedure with declared pressure and temperature limits.
3. A controller follows the procedure while sensors compare actual readings with the target.
4. An interlock can stop heat or feed when a limit is reached, but it cannot make the vessel’s accumulated heat, pressure, gas, or charge disappear.
5. The result, residue, and vented material are assayed or otherwise inspected before the procedure is treated as reusable.

This workflow makes a saved procedure an experimental claim, not an unlock. A procedure may be valid for one vessel geometry, feed mass, and heat source but fail when any of those conditions change. That is useful information to carry into the next run.

3. Quality-aware circulation and recycling

As soon as material has grade, purity, or provenance, an automated system can do more than move it. It can keep unlike batches from being treated as interchangeable.

For example, a planned beneficiation and smelting area can send a sufficiently high-grade ore batch directly toward roasting, send lower-grade material back through separation, reserve high-purity copper for electrical work, and route off-spec product into a reprocessing buffer. The decision must be based on an available assay or instrument reading and a player-declared threshold. It must not use a hidden exact composition simply because the simulation has one.

The resulting workflow has three benefits:

• It prevents a valuable pure batch from being diluted accidentally.
• It gives by-products and off-spec material a destination instead of a delete-on-overflow rule.
• It turns quality data into a real production decision, while preserving the uncertainty and sample cost that produced that data.

4. Hazard isolation and remote service

Automation is most valuable when it keeps the player from repeating a dangerous task. A later system can isolate a gas-producing apparatus, monitor its pressure or atmosphere, shut specific actuators, and move small loads through a hazardous area. At the electricity tier, drones are planned for remote, awkward, or unsafe jobs; they remain power-hungry and throughput-poor compared with fixed solid transport.

The safe workflow is not “hazard solved.” It is: detect the condition, isolate a physical path, account for the contained or released material, verify the atmosphere or apparatus state, then perform a declared recovery procedure. Remote operation prevents needless exposure, but it must not grant hidden information or erase the contamination, heat, pressure, or damaged material the player must still manage.

5. Grow-stack and life-support control

The later hydroponics system extends the same logic to biology. A grow stack is a bounded chemical domain: nutrient solution, pH, conductivity, light energy, water, carbon dioxide, oxygen, heat, and biomass all have observable roles. Automation can meter nutrient solution, hold a measured environmental range, schedule lighting, and route suitably scrubbed carbon dioxide from industry.

It should not reduce plant growth to a timer or turn exhaust directly into food. Selective ion uptake changes the solution; inadequate light or water changes the biomass outcome; contaminants in a flue stream must be handled before it becomes a plant input. The useful loop is an instrumented closed loop where the same skills used for a furnace — measure, compare, adjust, and account — become life support.

The potential is broad, but never generic

Almost every repeated workflow can become an automation opportunity when it meets the threshold: beneficiation, heat treatment, feed handling, atmosphere management, quality sorting, recycling, power distribution, remote hazardous work, and eventually biological/environmental control. The implementation choice follows the physical problem:

• Use a machine when a physical transform changes particle size, density separation, field exposure, temperature, pressure, or chemical state.
• Use a belt, chute, pipe, or drone when the problem is transport, respecting the payload and energy constraints of that medium.
• Use a sensor/controller when the problem is holding a measured variable inside a declared operating envelope.
• Keep the task manual when the player is still learning what inputs, conditions, or outputs matter.

That keeps automation from flattening Voxamine into a conveyor puzzle. The player is building an instrumented physical system: a factory that reveals why it works, why it stopped, and what must change next.

See also

• What Voxamine is — the design pillars, including “automate what you understand.”
• The playthrough — the bronze-age workshop where automation first becomes worthwhile.
• Game architecture — system ownership, asynchronous work, and durable-state boundaries.
• In-world vessels — the existing physical apparatus bridge that later machines build on.
• Docs/MASTERPLAN.md §27.3–27.6, §30.4–30.5, §32, §36.6, and §42.