5Axis Machine Cutting HELMET / DAISHIN SEIKI CORPORATION

It's hardest for CSG systems that can't say no. If you use Autodesk Inventor, there are some geometries where the fillet generator just gives up and says the geometry is too complex to fillet. Because it's a user-driven program, it has that option. But a machining simulator does not. Those are programs which simulate what happens as a tool cuts through a part, starting with the geometry of the blank and subtracting every cut. For 2.5D machining (tool always approaches material from straight down), the simulator can use a height field, and that works fine. But the general 3D case, where the tool can come in from any direction, seems to be beyond the state of the art in CSG to get perfectly. I've had both InventorCam and SprutCam blow up. If you spend $22,000 for a copy of HyperCam, it seems to work much better. this video Now that's a triumph of CSG.

What "original" means in the art world is amusing. It's all about monetization.[1] Cranking out a copy of Michaelangelo's David as a bronze would be a nice project. Hypermill/Daishin should take that on. Each year they do some insanely elaborate piece of art in metal as a CNC machining demo.[2] [1] link [2] this video

Carbide3D is still at "preorder". PocketNC has a very nice 5 axis machine for about $3K, but they were demoing two years ago and still aren't shipping. Roland has some nice machines starting around $5000, and those are real, shipping products. I'd really like to see PocketNC deliver. Milling software remains a problem. For 3-axis machines and simple work, there's the Cut2D/Vcarve/Aspire line, which is easy to use. Most work at TechShop uses that program. At the high end, there's Hypermill - $22,000 a seat, but you get your money's worth.[1] Look at the videos of it controlling a 5-axis machine, with long thin tools working deep through narrow openings, without screwing up. In the middle, there's SprutCam, from Russia, which is startup-grade software - clever, but buggy. (I've used it.) Since in the CNC business, bad software results in rejected parts and tool and machine damage, buyers are unwilling to buy bad software. There's some open source software for CAM, but it's below Cut2D, which is considered entry level. [1] this video

It's definitely possible to use a CNC machine to mill the shape of a rocket engine. You can do some pretty crazy internal milling with a 5-axis machine: this video But the problem isn't the shape of the bell at all. It's that the wall of the bell is not solid, but hollow, filled with plumbing. link The fuel actually flows through the pipes and voids in the red-hot engine, where it absorbs the heat that would otherwise melt the engine. It then flows back up to be burned. You'd need an endoscope, not a robot arm, to mill that out! And neither an endoscope (obviously) nor a robot arm would likely be rigid enough to mill in these materials. Robot arms are typically only used when milling wood, plastic, or soft modeling materials. A CNC mill in metals like the exotic Nobium alloys used in these engines needs incredible rigidity which robot arms just don't have. Consider the hardened tool-steel shafts of the end mills used in the first video - and those are easily broken, while just cutting aluminum!

There's also machining a piece of metal in an incredibly complex way, but with the right automation can be turned into a comparatively simple process. If you showed someone 100 years ago what we could do with a block of metal and a 5-axis CNC machine it would blow their mind. That's the type of area where automation can do great things. Look for places where "simple for a machine" and "simple for a human" diverge. I could do the job of every logic gate in my computer, but I wouldn't want to try and boot a copy of Windows with a pad of paper and a pencil. As an example, not something you'd want to make on a hand operated milling machine: this video But to look at something that's easy to do for people and think "That's easy, a machine could do it!" It's always going to depend on why it's easy. Handling situations that differ from one job to the next is a pretty strong point for humans, especially when those differences are handled by grabbing and moving things. But then, five years ago we would've put "playing Go" on there with grabbing and moving things, so you never know.

link specifically mentions a high-resolution jig. I wouldn't be surprised if in some cases just the tooling bits are around $5000, leaving aside the whole machine. For high-resolution machining, e.g. for lens fabrication, there's a whole science to drill bits, where special processes allow for achieving up to sub-nanometer surface roughness in parts. If they're making holes in hard metals they might be using titanium, diamond, or carbide drill bits that have crazy physical properties and cost a pretty penny. In some sense CNC machining is also a WWII era technology, but CNC machines from that era couldn't touch what modern-day machines can do and the repeatable precision with which they can do them: this video While on the subject, this really is a fascinating field and it rules the world more than people might think. Consider that TSMC's core machines are made by ASML. Among the most important and specialized-IP parts in ASML machines are the huge lens which are made by Zeiss. Zeiss makes the lens with some next level machining equipment from companies around Heidelberg and Karlsruhe area. High-resolution machining isn't super known to the general public but it is so fascinating once you get to see the things they can pull off.