The lack of at least weekly engagement by upper management and others in the chain is saddening. I get that Tim and Team have only one physical body each, but I've got to bleat on again that layers and other things need a massive, immediate, attention-seeking overhaul in order to lure new purchasing users. It seems as if forces want Punch SharkCAD to die in favor of other internal brands or apps. Ideally I'd be wrong on this assumption/fear/lament.

Although I'm not crazy about the AI+LLM frenzy, I would love to see an overhaul of SharkCAD and the issues I've whinged about be automatically performed by an AI or LLM code builder or code cleaner.

Ideally, someone will make an AI tool that is smart enough to make its own libraries, unemcumbers people and companies of stupid patents, and, crucially, enables companies and end users to just on the fly make whatever tools they need or want — and NOT be bound by/tethered to the dictates of any one OS or any one platform/hardware. Particular, I deeply despise apps that are just plain flat without visual relief. And I despise apps that feel too much like a phone app instead of a traditional desktop app. I don't like flyouts that feel like they dictate/constrain rather than facilitate.

At SOME point, if presumably successful app developers of any app can't or won't refresh or up-feature their offerings, then, given the frenetic pace of all things LLM based on what we've seen the recent two years of whirlwind AI+LLM heaped over the transom (and deer-gazing-into-oncoming-headlights ambushing and utterly destroying the unprepared), CAD is going to be one of those environments where massive consolidation +regurgitation will reshape the landscape of hobbyist and professional CAD andy anything in-between engagement.

I think AI+LLM+RAD in CAD should be more anout upscaling, up sizing, and ip-fraturing the actual creation side of the app, not its reporting/lensing features.

There is on YouTube a Julia McCoy who cloned herself so successfully it's enabling her to put out more content than she could singly do. I couldn't even tell I was watching her self-made clone. (I'm no expert clone-finder, and while I did know she was working on such, I forgot, and so I wasn't on guard or expecting to be pleasantly surprised she now can kick out more material to maintain her audience or whatever she aspires to gain in doing so.)

SharkCAD needs to see such in-house cloning so users don't lament that SharkCAD is dying or dead but we just haven't gotten the electric tickle.

Hopefully, Tim is cloning himself so his app remains real. When people make great things, such things should not be left to the dustbins of history nor to some board or PEF to dictate that something still quite useful dies because it's in the way of a new shiny thing. Perhaps SharkCAD needs a new, revolutionary APU and add-ons capability?

SharkCAD can be so much better if tweaked a little bit here and there and if the AI+LLM goes under the hood and stabilizes things and even brings realtime continuous, seamless, non disruptive backing up of files, too.

Or, is it that while this forum seems fading, there is a different communal/gathering place? I do keep forgetting, but, since I'm not part of any beta and can't afford to upgrade every year, I'm virtually in the dark. And, since I along with others can't seem to suggest anything that seems interesting enough for the company to implement, what ability or urgency have I (or others) to engage?

This vicious and virtuous circle and cycle will not be helpful.

I've only had 3 plays with AI.

In one i asked it to write Some CNC g.code
I asked it to cut a large square out of a certain sized sheet and to place a square pocket in a certain position on that sheet.
It was a bit eerie how fast it came back with some code saying, saying it should look like this but you need to customize it for your machine.
It took much less time, about 10 seconds, to write the 50 or so lines of code than it took for me to type in the 4 or 5 line question.
Did It work?
Well I don't write g code but after a quick look, it looked okay, however, I didn't use it. (I use a program that generates Code for my machine)

Can you possibly imagine how good machines are going to be in 10 years time when this stuff will be embedded into software?


I admire Tim for introducing AI into Viacad Shark. He seems to have been a trail blazer of innovative software all his life.


Ive also often wondered if AI is sufficiently advanced at the moment to analyze complex code.

It would be a great tool if Tim were to feed it the viacad code and ask it to analyze it for instances of the various bugs and to propose solutions to fix it.
I think this is one of AI's natural strong points.
Hackers are probably training it up on how to find weaknesses in security software as we speak.

However once he did that, would he retain the intellectual property or would the AI then add his programing knowledge into its general tool kit.

Edited by user Thursday, March 6, 2025 11:29:40 PM(UTC)  | Reason: Not specified

I use AI for a few hours every day, primarily for software and electronics design, troubleshooting, and searching through massive amounts of data. I’m amazed by how quickly AI has evolved over the past year—but also shocked by the sheer amount of mediocre and misleading output it generates. AI, as most people understand it, operates on a language-based understanding and isn’t necessarily technically correct. It lacks system-level comprehension and fails to account for environmental dependencies such as memory leaks, real-time system constraints, drift and tolerances, aging, thermal expansion, ESD, EMC, lightning surges, vibrations, resonance frequencies, water absorption, surface roughness impact on design, thermal radiation, and so on. It also doesn’t truly understand use cases, including how users or unforeseen events can deviate from expectations.

Additionally, AI lacks DFM (Design for Manufacturability) expertise and real-world experience. It’s not necessarily the best algorithm for solving certain problems (for example, FFT for frequency detection, even though a neural network could be trained for this purpose).

AI-generated designs are based on patterns learned from the work of average engineers (with IQs typically between 120-140, varying experience levels, and the occasional bad day when a design was created), as well as past AI-generated designs—which in some cases is highly problematic. As a result, AI rarely produces work that surpasses the average; it just does it faster. In my personal experience, AI generates vast amounts of output that could take a lifetime to verify, which is a major issue in safety-critical designs. It often makes bold claims and sounds highly convincing, even when incorrect. Many times, I’ve asked AI to verify its results, cross-check with sources, or provide references, only to find that the links it provides don’t actually support its claims. It’s only when confronted with concrete proof—like a specific parameter in a datasheet—that it acknowledges its mistakes. This is unsettling and, if undetected, could lead to serious consequences. I’ve spent a significant amount of time correcting AI-generated errors.

On a different note, I personally enjoy designing mechanical components, and I believe many ViaCAD/SharkCAD users—including artists, designers, and semi-professional hobbyists—would agree. AI-driven automation removes the creative aspect that makes designing enjoyable, making it a no-go for a program with this kind of user base. For these users, stability is far more important than AI features. Nobody wants their software to crash and wipe out hours or even weeks of work, especially in their valuable free time. The creative process is frustratingly disrupted when you have to redo everything from scratch.

Based on my experience, the development team behind this software seems relatively small, and it’s possible that much of the work is handled by a single person with support from remote teams. It can be that AI (which is new and exciting) has not been received as positively among users as expected, and that might have influenced development priorities. Regardless, I have deep respect for the work that has gone into this program, and despite its flaws, ViaCAD/SharkCAD remains a fantastic piece of software.

I sincerely hope the developer behind the software continues its development, and as long as updates remain reasonably stable, I’ll keep purchasing them.

Edited by user Sunday, March 9, 2025 2:50:27 AM(UTC)  | Reason: clarification