Jump to content

Leaderboard

Popular Content

Showing content with the highest reputation on 09/18/2026 in all areas

  1. I usually just keep it docked on the right side and use ctrl+1 to open it if needed. any of these will open properties (command "Properties") (command "PR") (command "CH") (command "MO")
    2 points
  2. I use a variation of the ssx.lsp to select objects in the drawing. In Bricscad v15 using (sssetfirst ss1 ss1) would both activate grips and activate the properties window to show info on those objects. For example select all blocks with same name, I could quickly edit the attribute value for all of them. I'm testing upgrade to v26 and the properties window no longer activates unless I manually toggle the property preview button. Is there a system variable to use the old behavior? Here is the simple code: (setq ss1 (ssget)) (sssetfirst ss1 ss1)
    1 point
  3. Honestly, I've been using AI to code in AutoLISP for the past year or so and I think it's more successful at it than I am. I'm not saying it's clean code, but I'm saying it's already useful. I had a sales engineer ask me info on a project that they're selling trying to verify if what they sold was feasible and with the tool the AI built, I could give them concrete details. If I had to finish that tool, it would have taken me months to write it. I had to test it a bunch of times in autocad and give feedback and it'll find the problem. I had to upgrade which models I used if it gets stuck or restart a new chat. I have a bunch of md files for it to remember how the codebase works but it's getting there. I would say the AI didn't code the whole thing, I still gave it most of the ideas to use, it just had to fill stuff in between for the implementation. At the end of the day, I just want it to work. I had it write a debug.txt which it can read after each test run. I had it create toy functions to measure what it did after the fact. There's a couple of good engineering practices still in place, it's just that I'm allowing myself to have 'fuzzy' details about the implementation.
    1 point
  4. @GLAVCVS thanks for the feedback. I assume to use a single Powershell session you would have to run a batch file that launches Powershell first before starting autocad? Then when autocad closes or crashes out that session of Powershell also closes.
    1 point
  5. I think AI is definitely part of the reason, but probably not the whole story. A lot of questions that used to become forum threads can now be answered quickly with ChatGPT or Copilot, especially when the problem is a small piece of code or a common error. The interesting part is exactly what you mentioned: when the AI-generated solution fails, people still need somewhere to ask a human who understands the bigger context. AI can produce working-looking code very quickly, but debugging someone else's real project is a different problem. I also think newer users are less attached to forums than they used to be. Many will try a search, ask an AI, or post in a Discord/Reddit community before they ever find a specialist forum. So the drop in posts doesn't necessarily mean people have stopped needing help; the questions may simply be going elsewhere.
    1 point
  6. Watching this whole "AI taking over" debate unfold, I feel like I'm seeing people in our industry fall into three different camps: First, there’s the Spectator. They use AI to generate flashy concept art or write basic emails. It’s pure consumption—cool to look at, but they don't really know how the engine works. Then you have the Skeptical Co-Pilot. This is where I sit for my daily development. You let AI do the heavy lifting for specific, tedious tasks, but you absolutely do not trust it blindly. You sandbox its output and rigorously test it before ever putting it into your actual workflow. Finally, there’s the Blind Believer. This is the whole "AI Agent" crowd trying to run a one-man firm. They type a prompt and expect a flawless final product the next morning. The reality? You have no idea what kind of spaghetti code or geometric garbage is hidden inside that black box, and auditing it takes longer than just doing the work yourself. Speaking of that "skeptical co-pilot" approach, I actually went down a massive rabbit hole with it recently. I was trying to solve a very pragmatic problem: How do we get an LLM to generate hyper-precise CAD geometry without eating up a million tokens and spitting out hallucinated garbage? Natural language just doesn't cut it for drafting. So, I worked with an AI to invent a prototype "Universal Morphological Matrix" (I just call it V-Code). The idea was to strip away all the human linguistic fluff and reduce CAD commands down to an absolute, 256-token hex-based root system. It basically compresses professional CAD knowledge into a tiny "Cosmic Language", which a local interpreter can then unpack back into precise AutoCAD/LISP commands. I pushed the experiment so far that I jokingly asked the AI to use this highly compressed logic to solve the P vs NP math problem. It confidently spit out a bunch of hex codes and claimed it had solved it. Did I believe it? Absolutely not, because I couldn't read its alien derivation track at all, haha! At the end of the day, AI is an incredibly powerful engine. But if you take your hands off the steering wheel and just blindly trust the machine, you're going to crash. It remains a tool exclusively for people who already know exactly what they are doing.
    1 point
×
×
  • Create New...