Pipelining The Factory of the Future
How do we make a factory better? Buy existing -> Build with Claude
Bidding on flight hardware requires Designing, Simulation, Prototyping, and Testing. Designing and simulation should be a snap of a finger (claude). Prototyping and Testing lead times could be shrunk with a fast enough qualification facility (protoshop + claude).
Designing and Simulation Solution - CAD and Ansys mixed with claude in a loop that will continue based on certain specification requirements and rules set by the engineer.
In code there is a system called linting which is basically stylistic changes, i.e. your company names variables a certain way and the linter will check all variables and auto update them. With enough hard coded barebones requirements and claude with access to CAD and analysis tools, the designing and simulation loop can be shrunk from 2-3 months to a week or less. Engineers will be moving from designing to designing to design.
Prototyping and Testing Solution - Automate shop floor and close loop for design.
AI scheduling MES through Warpflight that takes in everything in work orders, orders robots to go around and start different operations, automates writing the CAM for certain geometries, can check back with design+simulation system if certain items are not manufacturable and give feedback based on output of machines (cameras + automated CMM) that restart loop. Only a handful of machinists would be required to oversee this facility. It will be like one swe ruling over several claude codes. Testing will require more oversight from engineers for now but in house telemetry system will provide more visibility and more data for our systems to use moving forward which have unknown potential benefits when the system becomes more advanced.
The above was for the design cycle of new goods. Below is the private equity.
First thing would be introducing our own custom built ERP/MES (I know this one is trite) but the benefits are being on the cutting edge. It's doesn't make sense to buy existing companies and make their operating system the same. Buy new company -> migrate all data in 2 weeks hard deadline and stay on call for the next few. Core ERP/MES built out and proven in protoshop facility for less risk and pushed out in scheduled rollouts that are greenlighted by protoshop GM/supply chain to subsidiaries. Basically no slop pushing to subs, protoshop can be updated every minute but subsidiary changes happen after rigorous testing.
Then we move forward with our automations that we prove out at protoshop to the shop floor facilities. At protoshop we would have written several custom drivers + auto g code generators etc that are able to be used here. Protoshop as the ubermensch and subsidiaries that benefit from trickle down amcanomics.
More immediately actionable software things that interface with DLA workflow would be ingesting all the drawings/data from our subsidiaries and giving ai more context on precedent drawings to make a new design. Time saved from just amending older designs. Our advantage is all the data we've accumulated from the subs. Aiden had an example where he was able to leverage an existing ACPI drawing and make changes for a new product with time savings greater than 50%. (requires digitizing all info first).
These are spitballed ideas but the theme is the huge step function change here is a drop in cost and speed. Software is cheap and infinite but engineers are expensive and the real world is slow. Everything should be claude.