Post 1, intro/general

Post 2, flight enabled by plasma thrusters

Post 3, Immortality via brain-backups

Post 4, Global vactrain network

The Omni-factory is the biggest and most sophisticated type of factory in the world. It makes almost everything (including its own tooling). There are a few dozen to one hundred around the world.

It is fully-automated, human-free. In come shiploads of raw materials – and robots, computers, VR headsets, pharmaceuticals, and rocket-engines come out.

The factories are huge: 1km to 1.5km across and a few storeys high. Packed into that dense footprint is a station for every chemical, electronic, and industrial process.

Don’t make the mistake of thinking this is the only kind of factory in this world. There are localised means of production elsewhere, e.g. an iron refinery is sited next to a iron mine. Smaller factories, and more specialised ones, are in other places.

High-level superhuman AI only does top-level planning. The AMIs (advanced machine intelligence; also French for 'friends') write the code for the factory, and (dumber) controllers at each module handle the detail.

It all goes on the graph

'Graph theory' is a fancy way of saying 'dots connected by lines'. A 'directed graph' just means the lines are arrows. A 'directed acyclic graph' just means the arrows never form loops.

People talk about 'technology trees', but next time someone says that you can tell them in a nasally voice, "Ackshually it's not a tree, it's a technology directed acyclic graph". You know, to win friends.

Say you want to make a CHAIR... you take inputs like WOOD and METAL. The METAL must be made into SCREWS using a process called THREADING. The WOOD undergoes a process called SAWING. The SAWN WOOD and SCREWS go into a process called ASSEMBLY and the output is a CHAIR. We could illustrate this as a directed acyclic graph...

...where each ARROW is a well-understood industrial PROCESS, and each THING is a NODE. Ultimately it leads to treats (T).

Making an MRI machine or vactrain is no different from making a chair in this sense. The graph is bigger. How big?

• Limited number of inputs into each product – This video about digital communism says (around 51:30) an average 160 inputs to one workplace.
• Limited number of products in the world – https://commerce.net/how-many-products-are-there/ says "the QRS Catalogue is now up to 100 million entries (up from c. 80 million a few years ago)."
• Limited number of industrial processes – There's really only 8 ± 2 things you ever want to do with matter: take some away (grinding, abrasion), add bits on, mould/cast, bulk deformation (like rolling, stamping, or extrusion), cutting/sawing, joining, and coating. https://en.wikipedia.org/wiki/Industrial_processes lists about 70. Each of these can happen at various scales (making a tanker full of solvent versus a test-tube of a pharmaceutical). Various versions for various materials (cutting wood versus cutting metal). And with various precisions (a bicycle seat can deviate by 1mm and that's no problem, but a microchip needs nanometre accuracy). But all-in-all I think 400-1500 stations will cover everything even in a very advanced economy.

There is no astronomical complexity. There are less than 300 inputs per product, 400-1500 processing stations, and fewer than 200 million products. A computer (not really an A.I.) can easily plan a graph of this complexity, then it issues commands for robots to execute them.

The robots know where the goods are and what to do with them. "Robot 576, carry 250g of hydrofluoric acid to etching station #5, go!"

Industrial Processes are the arrows on the graph

Each arrow on the making-graph is a well-understood industrial process. In the Omni-Factory, each has its own station/module.

Examples of stations –

• Electric arc furnace
• Fischer–Tropsch process
• Polymerisation
• Haber process
• Tempering
• Work hardening
• Laser cutting
• Baking (ceramics)
• Electroplating
• Casting into moulds
• Biomineralization
• Growing mycelium into specific shapes
• Stamping
• Hole punching
• 3-D printing is worth talking about because when people think about futuristic manufacturing they always think about 3-D printing. The weakness of 3-D printing is speed and throughput; its strength is that you get complexity for free in one step. 3-D printing at 100 nanometres has been demonstrated; that's one 10,000th of a millimetre. The Omni-factory includes printers that can achieve that precision with metals and polymers, and multiple materials in the same print.
• CNC micromachining is similar to 3-D printing: weak on speed, strong on complexity
• Laser sintering is similar to 3-D printing: weak on speed, strong on complexity
• Photolithography
• Surface finishing units like tumble polishing
• Chemistry: Various molecular sieves, MOFs and nanoporous membranes tailored to specific molecules: compared to distillation, these can perform chemical separation with 10–100× less energy. Compared to Terra, the production of pharmaceuticals and specialty chemicals tend to rely more on solvent-free and solid-state reactions that proceed at ambient/low temperatures.
• Biology: bioreactors growing GM yeasts to make pharmaceutical precursors and bioplastics. Any sufficiently advanced production should integrate biology with machinery.
• Not smelting – that's done near the mines. Why ship ore to a factory?

...and each of these for different materials, different scales, different tolerances of purity and precision.

These stations will usually make big batchs, but the system is flexible enough to make one-offs when required.

Additive-manufacturing prints an injection mould. The mould is shuttled to the next module, where a production-run of 800 parts are made using injection moulding and stashed for later use. The mould is then melted down and made into a new mould.

Stations are sited rationally to minimise travel paths: after you make a silicon wafer, you need to etch it, so the etching station is next to the wafer-making station. But temperature-based co-location is a bigger consideration.

Traingang always talks about how dense cities are more efficient because they allow short bike-trips. Efficiently moving children to school is the same as efficiently moving ferrofluid rotary seals to centifuge assembly robots. You want things dense, co-located, with short trips. Products are moved around the omni-factory by gantries, pneumatic tubes, and wheeled robots.

Everything gets built far faster because components aren't being trucked around. Inputs move to the next step of the assembly graph in less than two minutes. Think how much of the manufacturing time of a laptop in Terra-2025 is hauling parts.

Inventory is the blobs on the graph

The global planned economy includes an exhaustive digital inventory. Every screw, every wafer, every vial of every chemical is tagged and tracked.

The inventory is global and includes the waste-stream. The Omni-factory ships your VR headset missing one screw and one panel, because it knows that they are sitting in your hometown from something someone threw out (robot garbage-men collected and tagged it). The headset is shipped to your town, the panel and the screw are added at the last minute, then it comes to your door.

Processing stations include systems to reclaim waste products. For example, if a piece of metal is milled, the metallic dust is swept up and added to inventory. Some waste products are biomined. Circular economy.

The AI can detect trends in requests, and anticipate demand. AI orders new factory-modules to be built to shift production from bulk polymer precursors to a pharmaceutical intermediate, anticipating demand. Processes are sunsetted as they become technologically obsolete or out-of-fashion.

Transport and robots

Imagine a room of stacked high with racks of boxes, screws, and panels. Gantry robots zip boxes from here to there. Vertical stacks of metal-organic frameworks and chemical feedstocks are marked with QR codes.

Robots (with wheels; no bipeds here), pneumatic tubes, and ziplines shuttle ingredients along the assembly line, with short distances from module to module.

Pipes are the best way to move fluids around at scale. There are pipes for oxygen (going from the cryogenic distillation wing where the oxygen's made to places like the hydrogen peroxide manufacture station), pipes for ammonia, for methanol, ethanol, deionized water type I, deionized water type II, etc. In-pipe robots inspect and the pipes.

Imagine a robot arm replacing catalyst cartridges in a chemical processing module

Robotic infrastructure includes: charging docks, spare-part storage. Robots and parts thereof can be made on-site.

A wheeled robot goes into the storage closet and swaps his screwdriver for a soldering iron.

• Inspection systems: Robots and embedded sensor networks continuously inspect every machine, every tool, every production-module for faults and cracks, and execute predictive maintenance routines. There are micro-robots for inspection and fine repairs. Distributed sensors for anomaly-detection. They might pause production when a fault is detected and you’ll get your treat a day late. Or an alternative production route might be found in the directed productive graph (technology tree) – for example the mould-injection system is temporarily offline so a part is 3-D printed instead.
• Containment & clean zones: graded cleanrooms (ISO levels) and isolated hazardous-process cells with active negative-pressure containment.
• Surfaces are made from advanced materials that use things like anti-corrosive and hydrophobic finishes to prevent corrosion and make spills easy to clean up.
• Emergency shutdowns with isolation valves and fire-doors.
• Safety & containment: blast-suppression zones around volatile processes, fire suppression using inert gases, automated venting and scrubbers for chemical incidents.

Temperature management

Think about Terran industry: in one place a furnace is hitting 2000°, and then that heat just leaks out of the building. Down the street there’s an chemical process that requires temperatures of 120°, and has to make its own heat from scratch! What a waste!

The Omni-factory is logically organised, with the hottest processing-modules at the centre (near the fusion power plant), medium-temperature processes in the middle, and room-temperature processes at the rim. People think of industrial processes as requiring huge temperatures, but there are lots of industrial applications of being just warm; "About 30 percent of industrial heat demand in Europe is below 100 °C".

There is a cold wing of the factory where processes like cryogenic distillation are all located. There are ways of making heat into cold (like the Einstein fridge or heat-pump).

Ocean site

Most Omni-Factories are built at sea (a few at the shore). Why?

• Shipping lanes
• Keep the noise pollution away from humans
• Dust is terrible for factories. Where does dust come from? Animals, plants, and soil – all things that don't exist at sea.
• Water is purified (via microbial desalination) and used in industrial processes. Water is the most important industrial chemical.
• Fusion plants extract deuterium from seawater, using graphene 1, 2.

Fusion provides the electricity for the factory. Geothermal can additionally supply hot steam and process heat. Also, putting the factory at sea provides an interesting look –