Yeah I ran into a ton of fillet issues, that’s what I’m referring to when I say I layered on a lot of my own kernel fixes. Mostly fillet issues and shell issues. It’s MUCH better than it was on pure OCCT. With how capable Claude is, I could’ve started from scratch, but it was easy enough to added fixes where necessary.
I’ll post the link here so anybody can give it a try. I’d love some honest feedback. There is a discussion page on the site for feedback
One thing that could still use a lot of testing is the surfacing tool set. I personally have never done surface modeling professionally so I’m kind of a terrible candidate for testing there.
For what it’s worth, I just spent about 20 minutes playing around in Solidcraft and the fillet tool there is significantly worse than mine. It gave me about a dozen different “not supported” messages and did become unresponsive at one point on a very simple part.
Also the sketch tools need a lot of workflow improvements like better automatic snapping, better automatic constraints, and quick line/curve switch
Like I said, give Leta a try and see what you think. You may be surprised.
It doesn’t, but I expect it will be ready in the next day or two.
The scripting is already built-in for it. It’s what Claude has used along the way to build hundreds of test parts to test every new and revised feature. Now I just need to make the script accessible for MCP use.
Last week I tried running it on a local 4b LLM, hoping that the average PC could pull it off, and it didn’t go very well, but I do want to try that again at some point.
I’m not sure that’s fair. The first FreeCAD release on Github was 11 years and that wasn’t the first. There still wouldn’t be a 1.0 if they rolled their own. And AI has absolutely changed the game. Sure there are things that could have been handled better but that’s true of basically everything.
Anyone recently canceled work related product/service subscriptions because some recently vibed clone (last ~1 yr) was good enough (for work)?
Curious how many of these vibed apps are actually being trusted/depended-on by people needing to pay bills.
Personally, I’m reluctantly paying for adobe cloud, since I only use a subset of features, for a subset of their products, for a subset of the month. There seems to be more vibed options than I have time to evaluate, guess I need to use AI to review and recommend what to use (when it’s unable to just generate the desired output directly).
Well I was about to spend some quality time with FreeCAD finally but this all makes me think I need to give it another 6 months and see where things are at with all this.
The progress is astounding. Keeping up with new releases, abilities, packages, all of it is hard. Not to mention trying to learn something new, knowing there is a very real chance the software landscape is going to be undergoing rapid metamorphosis for the foreseeable future.
If you’re trying to wait for things to settle, you’ll be waiting a long time. Although, it doesn’t seem unreasonable to wait for the 26.3 release before diving into FreeCAD.
See, that’s how I felt when I wanted to update my early 2016 MPCNC and you just kept cranking out update after update faster than I could print them
I see what you’re saying and it’s not wrong…but…I do suspect that waiting and seeing what continues to be developed and supported is worthwhile with so many new vibe coded options for a lot of things popping up. Vibe coding makes it easy to crank something out…but actually supporting it and making it useful for users long term is another story. Even with open source there are plenty of projects that have just faded away due to lack of interest over time and get abandoned.
Creating something with vibe coding is fun. Supporting it…not so much. And I don’t really want to put a ton of time into learning a tool that may not be supported for more than a year or two.
The development curve is so much faster than my understand curve.
It’s not possible to keep up with new releases across a broad range of interests!
Whenever I look into something there’s just one vital (for me) feature missing.
That’s why I suspect that applications developed over a long period of time have a lot of bloat, but also why they please all of the people pretty much all of the time.
I feel as though this stuff is like a sugar hit at the moment. One that I am really wanting to experience, but at what cost to the skills that I already have? This is not to disparage you guys who are by and large building on skills you already have, and I am happily waiting for you to finish what you are doing so I can reap the benefits!
For now, I’m making stuff that I’ve sketched on a bit of offcut or a figurative napkin - translated that to CAD for no other reason that to “build it” in my head (I used to happily skip that step) then I build it using combination of hand tools and machines even though I could probably do the whole thing in not much more time using only machines! There’s still a tactile connection between my fingers and my brain which much of this is seeking to remove.
In our Photography Club world - there’s a lot of honesty involved. The use of AI in post processing is universal as it is at the time of capture of the image - but it’s currently generally not acceptable to use parts of an AI generated image (generative fill) that diminishes the craft of the photographer - I can merge images I have taken, or even get AI to do that, just not use AI generated replacements.
I’m probably thinking the same is true for all things - making the mechanical processes faster is a good thing, but somehow the actual creative outcome must not be allowed to turn into grey-scale, or some sort of beige median.
Tbh I don’t care too much about support for CAD, I’ll be happily fine to work with SolidWorks 2018 forever. It’s a tool, if the tool does what I need it too do then I’m good.
Only issues for me are bugs and OS support.
Basically what I need is a CAD program as stable as SolidWorks but on linux, preferably opensource because of any OS induced bugs in the future.
Ok I’m using FreeCAD from Oct 2026 and I have the code for my version downloaded at C:\old_junk\freecad_source_mine_new_2\. I’m experiencing this weird bug sometimes, described in weird_bug.md. Look at all the public FreeCAD commits since then and also look in the code to see how to fix it. Don’t update anything else, just fix my bug, then rebuild it. Make no mistakes.
I don’t believe this means that actual coding will die, but I have a grandson who is currently studying just that at a technical college after graduating from high school and completing a certificate in game design and animation.
What advice could any of you who work among this stuff, give to a young adult who has the kind of personality that allows him to sit endlessly in front of a screen typing?
It seems to me, that like all things these days, actual job experience is a pre-requisite to checking vibe coded work too.
Work has recently canceled 3 3rd party contracts because we were able to vibe code our own replacements. We get to pick what abilities the apps have and our data no longer leaves our system or has to rely on outside software being available.
I and two coworkers each took a quarter to get to MVP and can now tweak the apps to better fit what we want.
We still have full security, redundancy, and scaling. The apps don’t have public access making them even more secure. And connections in/out are faster.
My time does cost money, but over time the app will slowly cost less and less.
As a SaaS company, we have to be careful and make sure our value to our customers includes things they can’t vibe code.
