Just checked out the screenshots... rather interesting, and quicker than doing the maths yourself. :)
Mike

Perhaps this is a dismal failure on my part, but, until today, I had not known of the existence of CADD Pro. Is this an outgrowth of SharkCAD, or a different product under the "family umbrella"? Ive heard of CADD as a term, decades ago, bit can't recall it in the context of a new app related to SharkCAD or ViaCAD or PunchCAD.

Before AI is something I'd go near in CAD/CADD (or ideally, the AI can do this on the fly for the user), I'd like to see the following occur, or know of the capabilities of CAD-AI nexii, among myriad other things:

— stretchable/resizable dialogue boxes to prevent layer names of the repeated kind end up being useless because the last few characters of the name are unreachable/inaccessible;

— snappable dialogues that confer new filtering and task-related history trees so the user can date-time tasks; make the rules set human readable and save-able as a session file that can be reused so crashes don't upend the user's state of mind or workflow;

— a real layer finder that includes a text string input field at each dialogue box with some kind if tagger so that complex finds and exclusins enable the user to be more productive when thousands of layers are involved in a file;

— more flexible geometry and associativity import/export facilities; it should be possible to import and export layers merely by copying and pasting a selected tree or branch without the paste action successively generation unneeded, distractingly long tip-level trees that break the readability/continuity of a users naming convention, and without losing associativity (export/import is the only way I know of to preserve geometry associativity,but it comes with the penalty of top-level tree/branch additions);

— resolve the Shift+Tab decreasing-zoom graphics issue; for me, in every instance of ViaCAD and SharkCAD I've ever had whether in Win7 natively or in Win run in VirtualBox, the act of doing Shift+Tab disruptively and distressingly messes wirh view zoom; the ONLY way to nix this vexatious and stressful behavior is to run two instances of Shark so that I can rapidly visually toggle between the active and the referenced files I'm using.

— permanently put an end to amnesiac tool palettes; if I have, say, 10 successively successfully engaged sessions (determined by a successful close and reopen that doesn't show a lattern of me hunting for the pallet.pos file or making new files related to palettes OR determined by a long but unclosed), non-crashed session), generate a set of time-stamped files (maybe up to 10 distinct subfolders to be fallbacks); these can confer the benefit of saving the user some hair, and could do double duty as session personality files; I lose hair when the app crashed and takes away my palettes and makes me have to rebuild or relocate them (imagine every time you oark your car, you find it in a new location upon needing it— it runs dine, but it requires playing whack-a-mole);

— add porcupines; the curvature circle just doesn't work for me — especially when trying to make compound curvature curves as in ship hull curves. If an AI is going to be useful in CAD/CADD it absolutely has to be able to look at user intent, as in, "What's the end goal of this user or of this geometry? Ah, a tank? A car? A ship hull? Thar be compound curvature for sure. He's added twisted wires slightly away from y=0, and can't close the model because of "all wires must be either all open or all closed", because the 3D kernel developer didn't anticipate, can't support, or refuses to support such til a later costlier tool chain can be foisted on the CAD app dev or that dev's customers..."...;

— I can never make reasonable use of the curvature circle because it scales out to like 10 to 100x the size of my ship model, and at such a zoom or scale level, it completely takes me out of the frame of mind, and puts me into, oh hell, I'll put in a disclosure "My dear customer/user, that's your homework, not mine, to do. I'm not that skilled a user to create a perfectly faired curve... Sorry".

— the constraints tools should not insist that geometry be planar. Sometimes, 3D objects are all there is, and in and if themselves can be usefully internally transformed to achieve helping the user achieve something.

— will an AI embed help the user give rotation to parts? Will its scripting engine let me build a grid that say, lets me simulate transit along a distance? Say I want the propeller to turn x nbr of rpms, and I want to estimate the volume/mass/viscosity of water moved, snd want to translate that into mass of the ship moved, then suddenly shif the rudder or alter the blade pitch between port and starboard, which alters the mass-inertia-direction components. I realize this is a domain issue for naval architecture apps, but lowly sub-50-MB FREE!Ship and FREE!Ship Plus (I'm speaking of the v 3.5 or pre year 2015, abandoned version, not the picked-up, GitHub-sited version) sort of does some of that to calculate turning circle, acceleration, and other things such as righting moments. (I also realize that if Tim's AI can do that, it might hurl the app into ITARS territory and, well, that's a whole other can of worms...)


— if I trial or join the v15 beta (assuming I'm even invited), will I find that my propeller blades created at X=0 via the polar array tool don't need to be copied and pasted to prevent their flying apart along a virtual ray as I displace the polar-arrayed/duplicated blades to x= 155 meters?


I presume or think AI needs to look at reasonableness, not just speed or anticipation of easy transforms.

It would be nice of the mentioned scripting is along the lines of what's in Grasshopper, Orca3D, Nemo, Nautilus, etc. If I have to choose between Rhino+Orca 3D, or the other add-ons, and the Shark or Rhino+add-ons will cost $1,000-$2000, then, as mcluchbas I don't want an Autodesk-looking GUI, I might be forced there.

Also, a useful AI needs to do something about layers duplication. Duplicating a part is more than duplicating it without a paired or new layer. It needs to allow the user to tear off and mix and match layers WITH and irrespective of the relevant geometry. If I want to create a proxy gas turbine engine of some 24 or 48 wires but mirror and rotate it, and the initial work is starboard, but I want to create a parallel layer in or outside of the tree or branch, "duplicate part" is insufficient if the org structure is flattened in any way. An AI needs to think like a human in those small but useful ways, but need not be attempting sentience.

— will the AI be private? I can't see an individual beavering away on a self-styled skunk works even for a second accepting that an external party is going to impose, suggest, or ask for privileged information of or access into a user's file.

This is particularly important to me because I'm a slow mover. 90% of the people who or MLs that would gain privileged access to my files wouldn't hesitate to run off with my arc out of frustration with or intolerance at waiting for me. In other words, I have a severe allergy to helping someone "beat me at my own game" when I'm hoping to feed myself via my own designs.

I'd almost rather be dead than let someone 100x more krganized and faster than I am just help themselves (I'm referring to people or systems way higher above Tim, not saying Tim's after my files) to my private activities, mask their access history, then recompose my efforts as if theirs, then block me from being able to claim my own work as mine. Such has happened before, and it's nefarious, not reasonable competition. At least, that's my perspective. It's why I still use Lotus SmartSuite Millennium Edition, because there's no "home to call out to" (as long as I keep it offline since jerks out there can creat stub sites hoping to ensnare someone whose year-2003 install may be reaching out to dead links), and why I operate my machines offline or strive to lawfully achieve offline activation of any new apps. Yep, I despise most apps that are demanding always-online use and that kill off local private use.

The world is not as rosy as we'd prefer. Trust is easily abused. Massive overkill of IP creation is my self defense from being defenestration from a future company I want to create. (Yes, I admit: I'm paranoid.)

Great. Please also double check "at positioned". It was also not working in 1652.

Cheers

What I really like about the Mass Properties feature is that I can obtain massing about the deck plates, bulkheads, stiffening and more.

Perhaps interactive AI in SharkCAD can help head off poor technique issues caused by the user?

I occasionally get some nasty geometry in convergence area of rhe model such as where the sonar head's aft neck meets the keel at keel baseline. Since it's a hit-or-miss thing, I sometimes simply cannot thicken the hull. If I can thicken it, I often can't shell it. It took years to get a keel skeg to be really thin. It's a beast to work through in solitude under the paranoia of being outpaced for asking for help.

(Of course, when I sell the hull model, I'll provide multiple versions of samples so my end users can forge their own improvements forward. Naval architects will often say "use fewer than 20 curves 15 even, or as few as you can get away with. Well, for a DDG-51-like hull to look better than what's in industry papers, and to make a hull that is better than the non-representatice 5415, and closer to or better than the ~5512 geosim model, I had to use some 8-10 curves in the bow region, ~15 between the bow and after body, and some 15 for the aft cut-up or run area that includes the skeg, and out to the transom edge. This is because I don't like making hulls via "parches". My sonar head and the keel skeg are organically all one piece. There all myriad disadvantages, of course. One disadvantage is some waviness along the keel. To combat it, I have to add more input or shaping curves, which adds to the mathematical complexity, and makes fairing more intensive to achieve.

But, I get to brag that ViaCAD and SharkCAD enabled me to achieve it. Other CAD app users show off gooseneck and bulbous bows, but, generally, they don't have to deal with the vagaries of the bulb body transitioning downward beyond the keel baseline. I dare say their work content is not nearly as hard as mine, at least based on the approach I use.)


That forced me to use the project-surface-to-body tool to create solids within the good areas of my hull, then slice the solids into individual zones or compartments.

The tradeoff led to an advantage (albeit at generation/s of extra geometry) of having a solid for the volumes, and a solid which I shell to obtain the local massing info of the hull zone itself. I separately build in the deck plates. The bulkheads are the shelled areas at the ends of a given zone. Nicely and fortunately, the shelling tool allows all sides to have individual thicknesses or zero thickness. Super nice.

The transom is super easy to cap off and close. The stem, however, is a serious amount of trial and error, compromise, and more. Since I use mirrored half-breadth curves as the basis for rebuilding a one-piece spline for each pseudo-station that is part of a network of some 30-50 shaping/input curves — to shape the hull — the forward-most curve cannot be closed (the ~"all wires must be either open or closed" thing). I tried trying to "cap" it off, usually to no avail. So, the curve in question is put at something like Y=0.00001m, to minimize whatever twice makes the capping tool fail.

Part of the tradeoff is once I get the hull how I want or can stand fiddling with, I move the hull inboard by -0.00001m, slice the surface, then mirror and join it, then add the transom and the rop deck. If lucky, I can get a stitch to become a shellable body.

An advantage to doing this is I can select and do "show only" for all the selected geometry without using clip planes. One thing I dislike about clip planes is the geometric center of invisible stuff still exerts a heavy hand on zooming, panning, and on selecting or snapping to objects.

(Also, another disadvantage to clip planes is they seem to be culpable in the creation of mouse-unselectable detritus, the stuff that looks like geometry, but can't be selected by mouse. To get rid of such stuff, I end up havding to select everything visible, hiding it, then hunting through any layer possibly containing it, then doing Ctrl+A, then hitting delete.

This is where another problem shows its head: the one non-resizable dialog box that makes it impossible to identify layers where deep nesting with similar names in "assemblies" or deep organization exists. So, if say, I route a duct and its parts through several layers, and I use clip planes to show or hide, then I move the clip planes around, somewhere in there is where in my case, the geometry is being virtually cut, then gets left behind. It could be days or weeks before I realize it was created, and it is tedious to hunt and remove.)

Hopefully, AI added to SharkCAD auto senses unexpectedly created geometry and flags the user before too much time has passed. This is a time where thought stream should be interrupted.

nope, as always, don't be...

Oh, belay my last/nevermind.

Just now, after numerous failed link-click attempts across two visits, the diwnload image dialogue suddenly displayed. Not sure why it failed earlier, since previously, I was also logged in, did view refresh, replied to a message/post, etc.

Here's something I've wondered more than once: Are these beta/WIP builds intended to be official versions for daily use? The fact that they're named beta/WIP, plus the warning on the MasterViaCAD page ("WIP and Beta builds have limited testing. Use at you own risk.") indicate that they are not general release versions.

Help -> Current Version has never once told me that an update is available. To find the latest builds one must:

1. Discover this forum.

2. Find the thread announcing the latest WIP build.

3. Go to the MasterViaCAD website and request an email using the form which warns that the build may be unstable.

4. Follow the link in the email.

Am I missing something? Are official service packs/point releases available somewhere?

I for one am completely sick and tired of all the software I'm using trying to find a way to integrate AI. There are some applications I can see it being useful, CAD is not one of them. I work in film and video, and all the software we use is trying to find ways to get marketing-compliant AI buzzwords stuffed into their new versions. The only example that has (so far) made any sense to me is the ability to do automatic subtitle generation from a video. And, predictably, it's about 60% accurate. You still end up having a to do a bunch of stuff to fix it.

Here's what I care about in my CAD software:

1) Stability.

Edited to clean up typos/grammar/spacing/formatting...

No words are added or replaced to alter meaning except as denoted in theee open and closed brackets [[[ ADDED; ]]].




Tim,

AI in SharkCAD would be immensely appealing to me and vastly improve my dejected state of mind if your Copilot can do the following;

— spreadsheet-agnostic, export the ENTIRE layer tree to a file that can be opened in as old as MS Excel 97, but not just Excel, but also be friendly to Libre Calc, and WPS Spreadsheets, and SoftMaker PlanMaker (so, no funny MS techno landmines; that company has too much a death-grip on data and user experience and on software devs);

— ability to regard the export as either of or both of a generic table or a BOM;

— treat a SharkCAD BOM as a snappable/reorganizable crosstab applet that can be linked to the drawing to drive what is visible or to reorg the Layer Manager itself;

— ability to analyze the exported tree/BOM for geometry statistics, volumes, solids, curves, and even recognize empty layers that appear to be functioning as administrative overhead by the user for lack of ability to do string searches (particularly since we're limited to single-letter jumps down the tree, something that forces me to use crazy symbols [[[ ADDED; sparsely-used first-letter prefixes... ]]] to sift through my thousands of layers) on layers;

— ability to enable the user to manipulate the BOM as a spreadsheet sheet's cross tab (within SharkCAD, whereas above, it would be a separate applet or product);

— ability to modify the BOM to add traits/flags/tags to the data structure and have the AI recognize that the user appears to be adding inside-of/outside-of the model skin/bounds;

— ability to import or reimport a sound/vetted, reorganized BOM without destroying the underlying database IDs that likely carry associativity tags [[[ ADDED; or app/applet-generated underlying code. ]]]


This would potentially make up for the seemingly hard-wired tree structure.

This kind of capability would allow users to — whether designing a house, car, plane, or ship — add, say a set of bounds or wrappers denoting:

— inside of model volume
— outside of model volume
— above model volume
— below model volume
— left/right/fwd of/aft of model
— altitude within model
— grid/zone/arbitrary-bounds-by-skin position by face selection
— tanks/bulkheads/floors/pipes/openings/stiffeners/etc (based on BOM tags).

Such would enable a mimic of the clipping planes, but not to hide things, but instead to help create arbitrary, resizable, lettered-numbered grids that can be BOM collection zones, updated in real time so the AI Copilot has user-driven bounds to steer the Copilot.

[[[ ADDED; the smart wrapper would infer by itself the fuselage, hull, car body, etc that constitutes the for-measurememts-purposes bounds of a body when sliders alter the position of a "measurement plane", a plane or bounds would be an enhanced clip plane (ideally a clip plane that doesn't generate non-mouse-selectable artifacts that I can only eliminate by first hiding everything that is Ctrl+A-selectable, then using Ctrl+A, then the delete key), said act/effort compelling systematically crawling up and down every suspect layer until all artificts are eliminated, said effort requiring unwanted and disruptive hide-unhide state changes — which can be numerous and impact session or later presentation intentions; I've had said artifacts return merely by my revisiting the clip plane tool, so, I've been deterred making too much use of the clip plane tool since the undesirable artifacts arrive without warning and without easy noticeability until well after the fact. ]]]

Such a Copilot could add a lot of value and set the stage for, say, adding or exposing scripting tools a la Grasshopper-esque features, but, say, germane to SharkCAD (so as to avoid litigation risk and to not excessively complicate the GUI/UI/UX).

While at it, this could be an opportune time to:

— make the Inspector resizable (my layer names are long to the right past the Inspector, rendering the inspector virtually useless in many situations, especially as there's no intelligent hot linking of the LM/CE/Inspector/geometry selection) [[[ ADDED; Select Mask ]]];

— make the Inspector, CE, Select Mask, and other tools snapped-filtered-smart so that tools not applicable to a given piece of geometry get dimmed out so the user's visual or cognitive load is cut by letting the eye focus on only pallet tools that will do something on or with a selected object or collection of objects;

— add onto the SM/CE/Inspector a database schema-like pipe/line/link thing that can down-select things in the drawing so that they're, say, made non-selectable more automated-ly based on what is in, say their string/drop-down boxes:


[[[ Added; include Layer Manager, too. ]]]

1. Show
2. Let be selectable
3. Dynamically hide if sensed as being repeatedly dismissed by or interfering with user.


In the case of a ship or plane, where compound or other curvature make make things ambiguous to decipher, the user might turn on opacity for the hull, but, when doing collision detection or interference checking, the users wants clearer boundary line visualization or adaptive non-opacity.


If the user can export the tree, analyze his/her work, logically deduce "I say let there be zones labled on these grids, and let me also say inside/outside/port/starboard/show me pipes/stiffening/bulkheads" so that auto tagging becomes AI-detectable, a user then could similarly add session areas of interest to reduce the amount of distraction when zooming/exploring the model.


And, please, if you deem all that above worthy of actually endowing in SharkCAD, please:

— allow layers in the Layer Manager to be dragged to anywhere except, of course, a circular or self-insertion kind, eliminating the "layers cannot be dragged outside of their..." message;

— eliminate the new-forced-inserted-top-layer thing.

[[[ ADDED CLARIFICATION; I'm referring to when a file is imported, said importation being the only (user-observable/accessible) way to preserve laboriously crafted associativity, said associativity not preserved by mere copy-and-paste action; on importation, one then must decide what to do about what visually comes off as a tattletale-like flag that a file import operation occurred — the new top-layer that needlessly clutters the view with the file name, disrupts tediously-crafted indentation and heirarchy intent.

A second consideration then is forced on the user: how and when to one-by-one remove imported empty layers, but not eliminate the "administrative-purposes" layers serving as labels for organization/reminders/notes...

This implies the AI needs to provide an auto flag, or the Dev Team needs to provide check boxes to help speed up eradicating, say, 2,000 imported layers that may be part of 15-50 sub branches/sublayers, said layers made empty because their contents in the source file were by user choice not selected for export, yet some of them are the administrative labels still needed in the new hosting model. ]]]


I hope I'm not the only one deeming these ideas worthy of being added to your baby. As I despise government saying "everyone's case is different, and no prior case sets precedent such that you don't have to prove your case as if new/never-explored", I strive to only ask for features that decidedly benefit more than myself regardless of model or designer injtial intent, such that users could quickly fall in love with my ideas.

Users, if my ideas above truly suck from a user perspective, I'm open to bashing, if you wish to upbraid me but, do so from a USER perspective, not a dev perspective.

0434-0511

Edited by user Monday, April 22, 2024 6:15:56 AM(UTC)  | Reason: Not specified