Hi again,

Another one for you folks…

PhysCAD

Another YT link from the developer

PhsCAD YT

Early days, but it’ll do CAM and other options they are working on, it’s a one time purchase

Mike

Hi ZeroLengthCurve



Yes, I thought about the vibe coding side too, but I’ve been guilty of using that in parts for the last two weeks, trying to get my DIY programmable car ignition of the ground. Over years, like my CAD dabbling, I have tried to get it over the line, and despite having a load of experience with commercial systems, I wanted something simple, that a small segment of classic car (Panhard) owners could relate too. I last looked at it in February 2026 and some Arduino code using libraries wouldn’t work for me when I was checking it on an oscilloscope, no matter what, so I laid it up.

My point is two weeks ago, I fed some of the code into AI, described the symptômes, and in an hour it had homed in on the error, and then went down a prompt/reply rabbit hole for 5 days on and off, before I called time and fed my simple working code from two years ago into it, and said why don’t you use a simple interrupt model as I only need this type of triggering sensor trigger….well it just got better. Then I got to the limits of the is microprocessor, so asked if as we aren’t using this processor specific library aka core engine, can we not use any processor know, which it could, and within a few hours it had done just that, but I got sent down another rabbit hole, before I pulled it back, and changed direction again. This type of AI interaction where you see patterns, and ask questions, query responses etc I have found truly remarkable,

I now realise this is the strength of AI for me, I can ask a specific question, say this and this is happening, it will claw through the code, and explain what is happening, and why, suggest a fix, and in no time you’ve got a solution. I just wish that SharkCAD developers used this, maybe they have or will, just a simple compare CAD kernels exercise would be huge, and it’s amazing these start ups (individuals who are passionate) have got this far, even using AI to find fault with NVIDiAs open source kerne, sending the results back to NVIDIA, who fixed it is huge.

Another thing you notice, is that they are multi platform dependant from the off, including a Linux port etc, and Apple Silicon, I can’t help when Tim sees these sort of developments, his cogs will be turning, or even maybe somebody in his team, will try and use it to solve their pipelines of development.

I hope so, Sharks are needed in the ecosystem.

Mike

Edited by user Monday, November 9, 2026 2:14:23 AM(UTC)  | Reason: Usual stuff typos, and autocorrection.

PhysCAD YT looks very promising. (Thanks @24c)
I think once you've watched this video, you'll stop testing other software.
I just bought it, and I think all the other investments in CAD software were a waste of money.

Edited by user Monday, November 9, 2026 5:47:02 AM(UTC)  | Reason: Not specified

I will also test PhysCAD, looks very promising. I tested Solidcraft, it opened large Step files quickly and the handling was very fast. But as soon I tried to move a connector, it hangs. So not very stable yet. Also it was not possible to copy & paste between files, very annoying!!

I love my Shark, but can't wait forever for a stable native Apple Silicon version and continue to pay upgrades. I stopped at v16 with MacOS 12.7.4. It works, quite stable, is ok but very very very slow if you group importer complex step files, almost unusable.

Let's see if this PhysCAD could be a good alternative. Is copy & paste between file available?

I started testing solidcraft wit the new kernel, very fast. However a simple task I frequently do in Shark on imported step file: just elect a hole and delete the inner face. Done, how gone. select a chamfer face, delete, gone. This is not supported on solidcreaft yet, very annoying.

Is this supported on PhysCAD?

The question is: solidcraft or PhysCAD? open source community driven vs a single man show?

If you could give some hints and experiences on both apps, would be great!

I hope the Company has its own AI agentic app-building servers and that it is clean-sheeting and internally adversarial-y recasting SharkCAD from the ground-up. (But, whatever it is, it cannot cause or cast doubt on human-made content or cause a human to fail to obtain copyright registration. In that regard, my v12 of SharkCAD lacking AI features is a bonus, unless it's true that v16/v17 only aid, not originate content.)

A whole new functionality refresh might be just what the company needs.

I suspect that all the major CAD companies that don't already have deeply-nested data integration features (such as databases that manage worldwide collaborative activity in a file or nest of files) already are doing this as a matter of due diligence to remain relevant.

It's either burn tokens on the Web, or burn them locally. But, if they don't burn tokens and reinvent themselves, one-person shops are going to burn the beyond recognition and off the playing board.

CAD companies no longer have the lizury of fixed schedule release cycles and forcing upgrades for access to no-brainer bug fixes. Just as MS is having to sitch Patch Tuesday, any worthwhile CAD company is going to HAVE to release bug fixes and feature enhancements weekly, and still find a way to offer perpetual and off-line product versions.

I wouldn't mind paying $20-$100 for major offline, out-of-band fixes and feature enhancements.

But, strangely, the bugs I see simply often aren't what others report, and what they do report often aren't things that affect me.

Alternatively, SharkCAD needs to have a Grasshopper-like offering so people can create an ecosystem for SharkCAD to break out of its current mould. I do ships, not airplanes. I don't do Rhino, even though it with Grasshopper and Orca/Nemo/Swordfish, etc, are powerful combos.

I don't need endless spline offerings, surface patches, and similar. I need intent-based spline smokthing that won't upset my columns/rows counts of nor the general positions of my spline control points. Why? In SharkCAD, my manually drawn splines look mostly ok.

But exported as .iges (IIRC) and imported into FreeCAD... They look horrid, to sngular, not spline-y. I don't know if people will interpret it as me being sloppy/lazy/elderly-shaky-handed, or as a function of .iges import into other apps. Yet, splines made organically in a given CAD app look far better than mine imported into FreeCAD. An algorithm that looks at the genrale and scale and network-grid obviousness could do winders for SharkCAD so a user has less tedium in trying to balance, say, a network of hull-shaping and hull-fairing splines. (This would apply to users who draw with curves, added-on-to with surfaces, then whose surfaces are stitched, then shelled, and still need to look smooth. People sculpting with meshes might have less (or different) concern about certain precision nit licks.)

I've bleated on about layers enough. A Grasshopper-like ecosystem that is open to users, and not just to vendors or special partners, is absolutely sorely needed and overdue.

There needs to be a new symbiosis. Even sharks have tagalong fish (and parasites) riding along, symbiotically. I haven't seen animal videos about sharks grooming/scratching each other up each other like gorillas picking ticks and fleas from one another's pelts and crevices, so, those hitchhikers must not be too bad to have swimming along in company.