AI doesn't solve the mature-CAD problem

Opinion

 

I read the maxi-tweet by Aaron Levie at https://twitter.com/levie/status/1776673128573284523. He's ceo at Box, a Dropbox competitor.

He notes this point in the discussion of the impact AI might (or might not, my words) have on workers becoming more efficient: In his example, 10 programmers, using AI, output the result of 15, so the company can now ship more features requested by customers.

This makes the software more desirable, leading to greater sales, which means the company has to hire more people (not fewer) to handle customer service, documentation, HR, and so on. The effect then spirals upwards, Levie figures, as the company desires to hire even more programmers who are 150% efficient thru AI.

Srini Chandrasekharan responded to say that Levie left out an important counterpoint: an inability of the market to necessarily absorb more product.

Which got me to thinking about the problem of matureCAD. The way I figure it, CAD finally got good enough about a decade ago. Depending on your drafting needs, CAD software even from 2004 probably does all you want. Me, I still use Visio 2002 and PaintShop Pro 6 from 2000, 'cause they do all that I need; newer versions of these programs overwhelm me with cluttered user interfaces trying to provide access to the new functions that I don't need. 

Beyond the basics of drawing and editing 2D entities and symbols (blocks), an adequate CAD system need only do this:

• Associative hatching and linetypes
• Associative dimensions
• A Properties panel
• Report areas, lengths, and counts -- BoM table generation
• Read and write DWG 2018 (the most recent format)
• Place images and PDFs in drawings
• Export in PDF and maybe a few editable vector formats

(Image source Back2CAD)

Sometime around 2015, big CAD vendors recognized that they could no longer deliver W-O-W! releases good enough to convince customers to pay to upgrade. By switching to the rental model, in which customers have to pay for the software over and over to use it by the month or a year at a time, CAD vendors no longer needed to work hard at adding new'n improved functions.

Less programming = fewer programmers -- all without needing AI!

(Another item Levie left out is that AI does a poor job of programming, and a good job of propagating bad code.)

So, I find it interesting that there actually are CAD vendors who offer the basics, some through Web browsers, such as Rayon (from Rayon; still being developed) and ARES Kudo (from Graebert). For the desktop, there is IntelliCAD with its permanent licensing starting at around US$380.

I feel there is still a market for basic CAD that the big CAD vendors, beholden to Wall Street, are forced to ignore. Perhaps, we'll see more fill-in products.

 

Comments

[Dairobi Paul]

Yes! The same goes for accounting software. MYOB, an Aussie accounting program, got to a stage in the mid-2000's where the additional upgrades provided no value to me, an owner of a small CAD company with a few part-time employees. The upgrades costed hundreds of dollars whose main advantage was updates on the tax tables and a few unnecessary bells and whistles. Like AutoCAD it went to an expensive subscription model. Having started decades ago using the manual double entry system, I have a pretty clear-eyed view of the overall simplicity of what accounting software needs to provide me. The simplicity of my mid-2000's perpetual license version allows me to adapt it to my Canadian book-keeping needs.