AI Coded Apps to Replace Mainstream Products?

[Dos_Box]

For most here I imagine that you have been aware that have not been living under a rock and have been reading up on the current furore over the vibe coded Artcraft suite and Photon, whose aim it to provide free alternatives (whilst looking very similar) to Adobe products. Artcraft even has its own Office 356 Word\Excel and PowerPoint equivalents. 
Both have been coded via AI, and both, after some playing with this weekend in a VM, are not there yet. Not by a long way.

However, they could well be with enough time, effort and tokens.

 

 

 

 

 

I'm sure there will be plenty of legal and ethical considerations flying around in the next few weeks\months, but the genie, I feel, is well and truly out of the bottle. And where will this ultimately lead?

Imagine a Linux based Office 365-alike setup, or tools that can mimic the functionality of Microsoft management and deployment tools and integrating with a current open source office suite?
Admittedly, there is a lot more to the software and tools we use daily than telling an AI to code you one, there are often hard learnt lessons baked in there that AI may ignore, leaving you wide open to lord-knows-what, as well as near-do-wells looking to exploit the gullible\desperate.

 

When it comes to interesting times, we are certainly entering them right now.

 

 

[krobin]

My whole YouTube feed is full of coverage from tech youtubers, designers, Adobe haters...
The things that impressed me most was how fast it came together. And now with tons of contributors (+ Claude/AI), this is going to exponentially get better fast.

Plus... the scale is ridiculous: they have done so many apps (12) at once! Adobe apps, Office apps, Autodesk, Pro Tools...

What software would you all like to see next? 😉

Here are a few:

 

 

[DalekSec]

Is scary how well Claude can build software, i simply asked it 'create a mac app that monitors my internet uptime'

 

and it did it, flawlessly adding in things i didnt think i needed but did, all within 20mins. Only thing i had to get it to do was change the icon as it was same as the WiFi symbol!

[2ilent8cho]

I wanted a native ARM Sim Tower for macOS, it decompiled it and made it in 2 hours. Spend a few hours yesterday enjoying it.  Also wanted to port a Windows 95 copy of Atomic Bomberman to macOS (which it never came out on) I now have that and it got to even swap out the IPX networking for multiplayer for TCP/IP. (I did buy those 2 games back in the 90s so not being totally naughty). Its reverse engineering/decompiling skills are very impressive. With how many game companies have killed games or just let them rot I'm all for it, as they won't think twice about screwing over the consumer, I expect this will be very disruptive. 

[machy]

I Think the effort here is where its going to fall down, they might be 90% of the way there, but its the last 10% where they will fall. After the publicity has fallen if they are going to bother to continue. like with most open source stuff they need someone to maintain it, which seems less likely with something that's been AI'ed overnight.

 

At least with Stuff Like GIMP your going to have someone maintaining it, or a fork of it until the heat death of the universe.

 

and with Adobe, £25~PP/Month for a business, or £4PP/Year for a school isn't that much really if you also take in your getting support.

 

 

i think AI is probably better for building modules/plugins for software rather than whole suites, in the long run.

 

 

Edit: That being said, the more it can be used for software preservation the better

Edited yesterday at 12:33 by machy

[Dos_Box]

52 minutes ago, machy said:

I Think the effort here is where its going to fall down, they might be 90% of the way there, but its the last 10% where they will fall. After the publicity has fallen if they are going to bother to continue. like with most open source stuff they need someone to maintain it, which seems less likely with something that's been AI'ed overnight.

 

At least with Stuff Like GIMP your going to have someone maintaining it, or a fork of it until the heat death of the universe.

 

and with Adobe, £25~PP/Month for a business, or £4PP/Year for a school isn't that much really if you also take in your getting support.

 

 

i think AI is probably better for building modules/plugins for software rather than whole suites, in the long run.

 

 

Edit: That being said, the more it can be used for software preservation the better

As regards business, Adobe currently have an offer on for 50% off on Creative Suite, down from £66.49 per month for the first three months. Then you get locked in.
Yes, it is decent pricing for education, but many in the private sector, particularly one-man-bands would consider the costs saving worth taking a punt with especially as the subscription goalposts keep moving, which does not generate goodwill from the end user.
By all accounts, the ArtCraft 'creator' did this as he was so frustrated with his Adobe subscription.
And never forget the way that some people will jump on any cause they think is worthwhile. And it's not that it's going to take that much of the volunteers time.
Just their tokens.

[nicholab]

My biggest annoyance is that in other areas of life, there is competition; why have we ended up here with software?  Why can't we agree that it drives the same way but has different source code?  Please don't bother doing this with SIMS!

[psydii]

The prompts that produced these apps apps are not being written by your average bloke picked at random of your local high street.

 

There is a very real chance that, certainly for a generation or so, there will be a transformation and democratisation of the software.

 

But after that the number of people who understand enough to work with a genAI to codevelop actually good products will diminish, and scarcity of skills will be dramatically worse than it was in the pre-genAI days.

Adobe might get unseated, but its replacement(s) will end up costing  just as much in 20-30 years when there aren't enough developers left to properly steer the AIs.

I do wonder whether  genAI will be able to overcome the problem of the systems eventually poisoning themselves after a decade or two of eating their own dogfood, which cold lead to the worst outcomes of all. Initial models that are good enough to replace a huge swathe of knowledge based skills, but then too limited by ingesting their own output to continue to drive things forward. Meanwhile people who could have steered the ship have aged out of the workforce with none/insufficient to replace them....  If I am wrong, then perhaps it will be good enough to write and maintain excellent code over decades without human supervision. 

[Dos_Box]

16 minutes ago, nicholab said:

My biggest annoyance is that in other areas of life, there is competition; why have we ended up here with software?  Why can't we agree that it drives the same way but has different source code?  Please don't bother doing this with SIMS!

You can guarantee that out there somewhere is a frustrated edtech trying to get AI to reverse engineer all of their subscription apps.
At best it will be cathartic, at worse the entire school might like a word with them.

[psydii]

"Have Claude, and the credit card. Let's see if I can code up an MIS before the school goes bust."

 

[Dos_Box]

5 minutes ago, psydii said:

The prompts that produced these apps apps are not being written by your average bloke picked at random of your local high street.

 

There is a very real chance that, certainly for a generation or so, there will be a transformation and democratisation of the software.

 

But after that the number of people who understand enough to work with a genAI to codevelop actually good products will diminish, and scarcity of skills will be dramatically worse than it was in the pre-genAI days.

Adobe might get unseated, but its replacement(s) will end up costing  just as much in 20-30 years when there aren't enough developers left to properly steer the AIs.

I do wonder whether  genAI will be able to overcome the problem of the systems eventually poisoning themselves after a decade or two of eating their own dogfood, which cold lead to the worst outcomes of all. Initial models that are good enough to replace a huge swathe of knowledge based skills, but then too limited by ingesting their own output to continue to drive things forward. Meanwhile people who could have steered the ship have aged out of the workforce with none/insufficient to replace them....  If I am wrong, then perhaps it will be good enough to write and maintain excellent code over decades without human supervision. 

It certainly is one thing that many don't appreciate about AI, in that you need to know what you are doing in the first place to judge the output.
AI is a great assistant, but a lousy creator.

[steve]

I do wonder if a lot of these will be flash in the pan. Without constant updates to address the latest vulnerabilities, you're going to find it hard to continue to meet any security standard like cyber essentials.

 

I'd expect Adobe will be heavily reviewing code to see if anything copyright has got into it too.

 

So might just be too much of a risk for organisations to take them on. Unless a government takes on some of these projects to get away from being reliant upon US software.

[MicraMan]

13 hours ago, steve said:

I'd expect Adobe will be heavily reviewing code to see if anything copyright has got into it too.

i doubt it if it is fully rewritten in rust...

[Dos_Box]

Interestingly, I was watching some US legal eagles on YT last night and one comment made me think. If Adobe had been using AI to code their products, does that create any possible IP issues on their part?
I can't see it myself, but Imagine if company 'A' uses AI to create a product via a third party AI, and company 'B' uses that same AI to create a similar product at the same time, utilising almost identical AI generated code and there are legal fisticuffs, who would be judged as having the copied\infringed the other?

[ICTDirect_Joe]

15 hours ago, steve said:

I do wonder if a lot of these will be flash in the pan. Without constant updates to address the latest vulnerabilities...

My thought exactly - what I'd really like to see is some of the pre-existing good open source projects to look at using the strengths of AI reverse engineering to improve and reiterate their own product, like LibreOffice and say Kdenlive, maybe GIMP. Build upon what they already have and potentially add features that previously seemed out of reach.

[Dos_Box]

Having been lurking on a few Linux forums over the past couple of months, there is definitely a great reluctance to use AI for coding in a lot of cases. There seem to be a lot of purists out there, which may cause some friction in the Linux world in the near future I think.

 

Saying that, I was having a play with some of their other offerings last night and there are some glaring bugs. In WordArt, fpr example, you can't change the font size. You can, but the second you click back on the doc it reverts to the default. That and a lot of other bugs too.
If you take a look at their GitHub page for the various apps, you will page upon page of reported issues, but they do seem to be working through them quickly (I image they set AI on them).

 

It'll be interesting to see what else starts to emerge over the next few months now people have had an opportunity to see the potential of AI coding.
*And yes, most will be horrible.