An anonymous developer has released Photon Studio, a free alternative to Photoshop. It opens Photoshop files, follows Photoshop’s tools, menus and shortcuts, and was reportedly built with AI for a few thousand dollars of usage. Reviewers say it is rough in places, but it works.
Most software companies never had a moat. They had a cost barrier. Copying an application properly used to need a skilled team and a lot of time, and almost nobody thought it was worth paying for. That cost protected plenty of products that had nothing else protecting them, and AI has removed it.
Your roadmap is now your competitor’s spec
Every feature you ship is now visible and cheap to take apart. Release notes, changelogs and launch posts read like specifications. A competitor’s agent can examine something your team spent months thinking through the week it goes live, and the gap between your launch and someone else’s version shrinks.
So being first with a feature buys less than it used to. The more useful question for a product team is what it can build that cannot be copied by looking at it.
What we found when we tried it
A free, open-source tool called REA (Reverse Engineer Anything) lets an AI agent take software apart. You point it at an app or a website, ask how it works, and it shows you the evidence. We pointed it at two things we were allowed to take apart: our own website, and OpenWhispr, an open-source dictation app whose licence permits it.
Working only from the outside, it worked out how our contact form screens out spam, down to the individual checks it runs before it accepts a message. It also found a bug we didn’t know about, traced it to a step in our own build process, and proved the cause. We have since fixed it.
On OpenWhispr, every specific claim we checked against the published source code was right, most of them to the exact line.
On our website, it could see what we had built but not why we built it that way. Its list of eight actions became three once someone who knows the system had read it. That still leaves a tool that can describe most of a product in detail, for free.
Is this you?
You had a cost barrier if:
- Your product’s value sits in the app on the user’s device, with little behind it on your servers (a desktop dictation tool, for example)
- Everything that makes your product work ships inside the app, so a copy of the app is a copy of the product (like Photoshop)
- Your pitch to investors rests on how hard your product would be to build, more than on your customers or your data
- You have built something on someone else’s AI model, and most of what you add runs on the user’s device
- What keeps customers is mostly the hassle of switching, rather than anything they would miss
You have a moat if:
- The valuable logic and data sit on your servers, out of reach
- Your product gets more useful as more people use it
- The detail your customers depend on took years of contact with the market to get right
- Your customers would stay if a cheaper clone appeared tomorrow
Is this you?
Where the moat moves
If you scored badly, keep shipping, and change what you are building towards.
Move the valuable work onto your servers. Anything that ships to your user can be taken apart: the code behind your website, your desktop app, your mobile app. The code that runs on your own servers never leaves them. A tool like REA can watch the requests an app sends, but it cannot see what happens when they arrive. Our own website showed both sides: REA described our contact form in detail and yet could not see what happens after you press send. A valuable process was invisible.
Build things that compound: data that gets better with use, a network that gets more useful as it grows, relationships with customers that a clone cannot inherit.
Invest in the detail. I used what REA found in OpenWhispr to build my own version of the one Wispr Flow feature I use every day: hold a key, speak, and the text appears. The main feature took little effort. I didn’t want to rebuild the dozens of small details that make the real one great to use. After a week on my own version, I went back to Wispr Flow, not perfect, but way better than the version I built from OpenWhispr. Those details take effort and time to polish, and they protect Wispr Flow. One person saving £20 a month (like me) had no reason to keep going.
However, detail alone won’t stop a funded competitor with a team, route to market and every reason to get the details right. It buys time, and it works best alongside the other two.
On the law: copyright protects the code itself, and many licences ban reverse engineering. We only use these tools on our own software, on open-source software whose licence allows it, or on a client’s software with their permission. I don’t expect everyone to be so careful, and I am curious to see how much of it turns out to be enforceable. Get proper legal advice on this.
The other side
The same tools cut the other way. If you pay heavily for software and use a narrow slice of it, rebuilding only that slice is now a realistic option for the first time. My dictation app shows where the limit is: not worth it for a £20 subscription, but worth a serious look for a system that costs you a lot and does far more than you need.
Either way, cloning your product is now fast and cheap. So the real question is: what are you building that cannot be cloned?
If you would like to talk through what your product would still have if it were cloned tomorrow, and how you can sharpen it to maintain its edge, I offer a free 30-minute diagnostic call with no obligation.
Fill in the form on our contact page.
