Design Is the Easy Part. Flux Is Built for the Rest.

Christian Keil wrote an essay this week with a title that every hardware engineer will understand: Superhuman Hardware Design Is the Easy Part. His argument is simple. AI models will soon design hardware very well. But the link between the digital world and the physical world is lossy. A design must still become parts that you can buy, files that a factory can build, and an assembly that fits together. Then the real world must tell you whether it works.

We agree. We see it every day at Flux. A good design is the start of the work, not the end.

We saw it first on our own team. As the Flux agent got better, designing a device became almost free. A board that took weeks took an afternoon. But the devices did not get to the bench faster. The time went to quotes, file checks, supplier emails, and order forms. Manufacturing became the friction.

That is why we started to invest in manufacturable designs many years ago: live parts, manufacturer rules, and checks on every design. More recently, we started to invest in getting the devices manufactured. Today the Flux agent gets the quotes and places the orders for you.

That is why Flux is not only a design model. Flux is the workspace where the design connects to everything after it: parts, factories, fit, and tests. The AI engineer in Flux designs the electronics and the mechanical parts of a device in one project, and it checks each design against the physical world before you order anything.

Parts that you can buy, when you need them

The essay tells a story from the first Falcon 1 launches. A test was delayed because one capacitor was in stock only at a store about 6,000 miles away. Every hardware team has its own version of this story.

When Flux selects parts, it checks live pricing and stock at distributors such as Digi-Key, Mouser, LCSC, and Arrow. Mechanical hardware comes from suppliers such as McMaster-Carr and MISUMI. Flux reads the datasheet of every part and checks the design against its pinouts, ratings, and limits. The bill of materials stays current as the design changes. The result is a device that you can buy parts for today.

Designs that your manufacturer can build

A design that is ideal on screen can still be hard to make. Design for manufacturing is where the essay expects AI to struggle most, because it needs knowledge of real machines and real processes.

In Flux, you define your layout rules and your manufacturer's capabilities once. Flux checks every design against them, every time. Electrical and manufacturing rule checks run on every schematic and board. When the design is ready, Flux gives you the files that your manufacturers need: Gerber, ODB++, IPC-2581, bill of materials, and pick-and-place for the board, and STEP, STL, or 3MF for the mechanical parts. Send them to fabricators such as JLCPCB, PCBWay, and SendCutSend.

From quote to order

A file that a factory can build is not yet a part in your hand. Somebody must get quotes, compare prices and lead times, place the order, answer the questions from the factory, and record the order. On most teams, that somebody is an engineer who should design the next device.

The Flux agent does this work for you. Ask it what your device costs to make, and it gets manufacturing quotes for each part of the device:

• Circuit boards: PCB fabrication and assembly, from manufacturers such as JLCPCB and PCBWay.
• 3D-printed parts: enclosures, brackets, and fixtures in resin, nylon, and metal.
• CNC-machined parts: aluminum housings, heat sinks, and mounts.
• Injection-molded parts: enclosures and covers in production volumes.
• Wiring harnesses: the cables that connect the boards and the parts of the device.

Flux reads the quantities, the board size, the layer count, and the materials from the project, so you do not fill in quote forms. The agent shows the quotes in one table, with the delivered cost and the lead time of each. Ask "What if we build 1,000 units?", and you get the new prices in the same chat.

When you choose a quote, the agent places the order for you. It also follows the order. Ask "Where are my boards?", and the agent tells you the status.

Your email and your PLM system, in the same loop

Much of manufacturing happens in email. A board house asks about a drill size. A supplier sends a new lead time. A molder sends the report on the first parts. Connect your email to Flux, and the agent helps you manage these messages. It reads the question, finds the answer in the project, and writes the reply.

The Flux agent also connects to your PLM system, for example Arena, and to the other systems that your company uses. The bill of materials, the revisions, and the released files in the PLM system stay the same as the design that you ordered.

Parts that fit together

Every part can meet its drawing, and the assembly can still fail. Tolerances stack. Teams that work in separate tools find the problem late, often when the first board does not fit in the first enclosure.

In Flux, the board, the enclosure, the brackets, and the wiring harnesses between boards are in one project. The agent checks fit and clearance on the assembled parts before it exports them. A connector that moves on the board is checked against the opening in the enclosure in the same project, not in a meeting two weeks later.

Simulate first, so the first build confirms

The essay makes a point that we like a lot: with better simulation, the physical build can be confirmatory, not experimental. Atoms stay in the loop. But each trip to the lab should confirm a design, not discover what is wrong with it.

Flux simulates and checks its own work before you build. It runs SPICE simulation of analog, power, and mixed-signal circuits, and full-wave electromagnetic simulation of the routed board. It runs thermal and airflow simulation of boards and enclosures, and finite-element stress analysis of brackets, enclosures, and mounts. The agent reads the results and changes the design. Problems show up on screen instead of on the bench.

Process knowledge stays with your team

The essay ends with process knowledge: the machinist who hears a problem in the CNC machine, and the engineer who knows why the last product failed in the field. This knowledge is rare, and it is usually in the heads of a few people.

Flux lets your team write it down once. Design guidelines, preferred parts, manufacturer rules, and lessons learned become team knowledge. Every AI engineer in your organization applies them on every project. Past decisions stay with the project, so the next design starts from what your team already learned.

Where this ends

When AI makes design easy for everyone, the bottleneck moves from designing things to making them. More teams will build more devices, and each device must still survive the parts market, the factory, the assembly line, and the test bench.

An AI that only designs is not enough for that world. Teams need an AI engineer that gets working hardware built, from the first idea to the order at the factory. That is what we build at Flux.

Describe the device you need. Flux designs the board and the mechanical parts, gets the quotes, and orders them.