AI and duplication of commercial software

We are hearing mostly about the risks of AI wreaking havoc on infrastructure or being used by hackers or other bad actors. Measures to prevent this are going to be difficult to implement given the race of the AI companies to make their products profitable and more powerful.

One thing that seems to be happening sooner is that it seems relatively easy to use Claude to rapidly make clone versions of commercial software without copying any code. The port of the Adobe Suite to Rust in a matter of a few weeks by a small software firm is the biggest example.

So before AI kills us all, it might kill a bunch of big software companies. If you can just point a bunch of agents at a software suite and tell them to duplicate the package and in weeks you have a duplicate suite with 80-90% of the functionality of the original, how long before you could have a 99% version or even a superior version?

I wonder if Anthropic is having a hard time selling more Claude to Adobe?

PS I haven’t heard about any current “look and feel” lawsuits and I am surprised that that issue hasn’t come up in the excited discussion of the Adobe suite open-source clones.

Code is cheap, but polish isn’t — yet.

Photoshop is, what, 20-30 years of built-up features, UX research, etc. It’s a lot of small and tiny things that add up to an enterprise-ready suite.

Long before AI, there was already Paint Shop Pro, Paint.net, GIMP, Photopea, etc. Another Photoshop clone doesn’t really change that. (And last I heard, people weren’t super happy with the clone yet… it’s superficially the same, but had numerous bugs and UX issues under the surface).

Let’s say the clone is 80% as good as real Photoshop now. If you spend enough tokens and human hours improving the clone, you might eventually get it to, say, 98%. But by then Adobe could’ve been doing the same and improving the real one to 110%. And the clone then just becomes another competitor doing actual software development, same as any other.

AI agents are pretty autonomous now, but one notable weak point that I’ve noticed so far is their UX/UI sense that still requires a lot of back and forth review with a human to come up with something good.

For now, in the industry, it seems like AI is mostly displacing junior developers while more senior ones are shepherding the agents. And big software companies are especially well-positioned to take advantage of them. If one person can create a Photoshop clone, what can Adobe itself, with 1000x the token budget and thousands more employees, do?

All that is to say… I think it will definitely affect big and small software companies alike, but the main threat isn’t from lone-wolf clones. It never was; they’ve always been around, they just don’t really have the resources to really polish something up to commercial standards. Software development has always taken advantage of efficiencies of scale that favor larger organizations, and that’s truer now than ever before, since it’s pretty much money in, code out. Up to a certain point, at least… at some scale you lose focus of engineering and just become a financial organization that’s not so user focused anymore.

I am, however, excited by the prospect this brings to startups. Some teenager in their room is gonna make the next trillion dollar company that enslaves us all.

That Rust adobe implementation is nice; the acrobat version runs better than whatever slop Adobe’s has become.

Making something that’s better than Adobe Acrobat is not exactly a high bar to clear.