Why code alone doesn't protect a software business anymore, my list of boring industries nobody is building for yet and the two prompts to find your gap in one.
In October 2026 one developer, Brandon Thomas, rebuilt seven Adobe apps and put them on GitHub for free. Notebookcheck reported he used Claude Opus 5.5 as his coding assistant. There's PhotoCraft for Photoshop, VectorCraft for Illustrator, FilmCraft for Premiere, LightCraft for Lightroom, a PDF app for Acrobat, EffectCraft for After Effects and DesignCraft for InDesign.
They're early and not every one is ready for daily work. But the PhotoCraft repo had more than 74,000 stars on GitHub when we checked on 29 October 2026. People clearly want them.
So if the screens of a giant like Adobe can be rebuilt by one person with AI, the screens of your app can too. Code on its own isn't a moat anymore.
Building another planner, to-do app or developer tool feels safe because you understand it. It's also the first thing everyone else builds with AI, so you're competing with thousands of copies on day one.
AI can rebuild any screen. What it isn't aware of yet are the rules and the history inside each industry. That's where the hard, valuable work lives.
I've spent 16 years building software for mortgage advisers, payments companies, property firms and compliance teams. None of them were exciting apps. The hard part was never the screens. It was the regulations, the old databases I had to migrate and the credit checks everything pulled into. That's exactly why those products are hard to copy.
Here's my list. Each one answers to real rules, runs on old systems and still does too much by hand. Pick one you know or one you can get close to.
1. Find the gap. Paste this into Claude with one industry from the list. It makes Claude show its sources, so you can check the rules yourself.
I want to find a software gap in this industry: [THE INDUSTRY, E.G. LETTINGS AGENTS IN THE UK] Act as my research partner. Tell me: THE RULES: which regulations and regulators this industry answers to and what they make firms prove. THE PAPERWORK: the jobs people still do by hand, in spreadsheets or in old systems to stay compliant. THE OLD SYSTEMS: the legacy software and data they're stuck with and what it connects to. THE PAIN: the three jobs that cost them the most time or the most risk if they get them wrong. Then give me three small software ideas that solve one of those jobs. For each, say who pays, what rule it helps them meet and what would make it hard to copy. Link your sources. If you're not sure a rule applies, say so. Don't guess.
2. Stress-test the idea. Paste this in the same chat once you've picked one of its ideas.
Here's my idea: [YOUR IDEA IN ONE LINE] for [THE INDUSTRY]. Be my toughest critic. Could a developer with AI rebuild the screens in a weekend? If yes, what would they still be missing: the rules, the data, the integrations, the trust or the approvals? List the five questions I should ask ten people in this industry before I build anything.
Then talk to real people. Ten conversations with people in that industry will teach you more than any prompt. The rules they complain about are your gap.
You've got the list and two prompts. Do one thing today. Pick the industry you know best, paste the first prompt and read what Claude finds.
What this won't do: it won't make a regulated business easy. These industries are slow to buy, they'll ask hard questions and some ideas need approvals or legal advice before you launch. Claude can also get a rule wrong, so check every source it gives you. That slow, careful work is exactly what makes your app hard to copy.
AI can rebuild any screen. It can't rebuild the rules.
Sources, opened 29 October 2026: Notebookcheck, Experienced Rust developer recreates Adobe suite with AI assistance as open source (8 October 2026); storytold/photocraft on GitHub; FCA, Consumer Duty. Rules change, so check the regulator's own site before you build.
The Audacity to Build is my community for people doing exactly that. It is opening soon. Join the waitlist and you are first in when it does.
If this was useful, follow me. I share what I am actually building, the tools that survive my own testing, and the ones I quietly cancel. New plays every week.
Pop in your name and email and the guide is yours, along with the occasional new guide when I publish one. No spam, ever.