Developer Who Cloned Photoshop With AI Just Did the Same to Microsoft Office

Last week, the internet caught a glimpse of what happens when a motivated developer stops treating LLMs like a search engine and starts treating them like an automated factory. Brandon Thomas, a developer who recently made waves by building a functional Photoshop clone using AI, has turned his sights on the most entrenched software suite in history: Microsoft Office.

Thomas used Anthropic's Claude to generate open-source versions of Word, Excel, and PowerPoint. They aren't perfect. They aren't going to replace the enterprise versions of these tools tomorrow. But they work, they were built in record time, and they cost almost nothing to produce. For anyone building in the crypto or AI space, this is a loud signal that the walls around proprietary software are thinner than we thought.

The Illusion of the Proprietary Moat

For decades, Microsoft, Adobe, and Salesforce have relied on a specific kind of moat: the sheer complexity of their codebases. The idea was that no startup could ever catch up because these tools represent millions of man-hours of legacy development. You couldn't just hire ten developers and build Excel from scratch; the math didn't work.

AI just changed the math. When a single developer can prompt a model to handle the heavy lifting of UI architecture, formula engines, and document rendering, the cost of replication drops by orders of magnitude. We are moving from an era of "moats built on features" to an era where the only real defense is brand, distribution, and live collaboration ecosystems.

Why This Matters for Builders

If you are a founder, you need to look at Thomas's project and ask yourself: "If a guy with an LLM can replicate my core feature set in a weekend, what am I actually selling?"

Too many teams in Web3 and AI are still trying to sell basic utility. They are building "the AI version of X" or "the decentralized version of Y." If your value proposition is just the functionality of the software, you are in a race to the bottom. Open-source clones will follow you like a shadow, fueled by the same models you use to build.

• Speed is no longer a differentiator: Everyone has the same access to high-velocity coding tools.
• Basic features are now commodities: Text editing, spreadsheet logic, and slide layouts are now library functions that AI can assemble on demand.
• User experience is the new battleground: The clones are rough around the edges. The winner is whoever makes the tool feel less like a utility and more like a partner.

The Open-Source Surge

What Thomas is doing isn't just a parlor trick; it's a demonstration of the new open-source reality. Historically, open-source projects lagged behind proprietary ones because they lacked the funding for massive UX and QA teams. Now, an LLM can act as a bridge, helping small teams or individuals handle the "boring" parts of development that usually require a headcount of fifty.

We are likely going to see a flood of "good enough" software. These are tools that satisfy 80% of a user's needs without the subscription fees, the bloatware, or the privacy concerns associated with Big Tech. In the crypto world, this aligns perfectly with the ethos of self-sovereignty. Why pay Microsoft a monthly fee to store your data on their servers when you can run a locally-hosted, AI-generated clone that saves directly to a decentralized storage network?

The Skeptic's Corner

I want to be clear: these clones are not "Office killers" yet. They lack the deep integrations, the security certifications, and the twenty years of bug fixes that make Excel the backbone of the global economy. If you try to run a Fortune 500 company on a weekend AI project, you are going to have a bad time.

However, the skepticism shouldn't be about the quality of the current code. It should be about the trajectory. If this is what one guy can do with Claude 3.5 Sonnet, imagine what a small team can do with the next generation of models. The gap between "rough clone" and "viable alternative" is shrinking every month.

Building for the Post-Software World

If software can be generated for pennies, the value shifts to the data, the network, and the trust. This is where crypto founders actually have an advantage, provided they stop focusing on the UI and start focusing on the protocol.

The code is becoming the commodity. The protocol and the community are the only things that can't be cloned by a prompt.

We are entering a phase where the "builder" isn't just the person writing the code, but the person who can orchestrate these models to solve specific, high-value problems. Brandon Thomas didn't just write code; he directed a system to build a product. That is the new founder skillset.

The Takeaway

Don't get distracted by the rough edges of these AI-generated tools. The significance isn't in how they look today, but in the fact that they exist at all. The barrier to entry for building complex software has been demolished. If your business model relies on charging for access to standard software features, start looking for a new model. The future belongs to those who build what AI can't replicate: community, trust, and unique hardware or data moats.

Read the original at Decrypt →