How One Developer Used AI to Clone Big Tech Software
A solo developer building open-source versions of Photoshop, Premiere Pro, Acrobat, Word, Excel, and PowerPoint would’ve sounded absurd a few years ago.
Now it sounds early.
That’s the point behind the viral story of Brandon Thomas and ArtCraft, an open-source project described in the video as a one-person effort powered by Claude AI and Rust. The apps aren’t polished replacements for Adobe Creative Cloud or Microsoft Office. They’re early alpha software. Some parts will be incomplete. Some will break.
But the important part isn’t that one developer built perfect clones. He didn’t.
The important part is that one developer could stand up working foundations for complex desktop apps at all, and fast enough for people to notice.
ArtCraft shows what AI coding agents are actually good at
The ArtCraft story is easy to overhype if you frame it as “AI replaced Adobe.” That’s not what happened.
A better read is this: AI helped one developer create a lot more software than one developer normally could.
According to the video, Brandon Thomas used Claude AI, specifically Claude Opus 5.5 as named in the summary, alongside Rust to build a suite of open-source desktop tools under the ArtCraft umbrella. The project includes creative apps modeled after well-known Adobe products:
PhotoCraft
A Photoshop-style image editor.
FilmCraft
A Premiere Pro-style video editor.
VectorCraft
An Illustrator-style vector design tool.
PdfCraft
An Acrobat-style PDF tool.
LightCraft
A Lightroom-style photo workflow app.
Then the scope grew. The project moved into productivity software too, with apps aimed at the core functions people expect from Word, Excel, and PowerPoint.
That’s a wild product map for a single developer.
To be clear, these kinds of apps are hard for a reason. A mature image editor needs layers, masks, color tools, file format support, canvas performance, GPU acceleration, undo systems, plug-ins, export settings, and a user interface that doesn’t fight the user. A video editor adds timelines, codecs, audio sync, effects, rendering, thumbnails, proxies, and huge performance demands. A spreadsheet brings formulas, charts, file compatibility, calculation engines, and edge cases that have built up over decades.
So no, AI didn’t magically erase the difficulty.
What it did was lower the cost of getting from nothing to “there’s a working skeleton here.”
That’s the real story.
The old software moat was built from time, people, and code
Big software companies have always had a few strong defenses.
One was distribution. If schools, agencies, studios, and companies already use a tool, switching is painful.
Another was file compatibility. If everyone sends `.psd`, `.docx`, `.xlsx`, or `.pdf` files, the tool that opens them best has a huge advantage.
Another was trust. Creative pros and businesses care about stability. Nobody wants to lose a client project because an alpha app crashed during export.
But the biggest moat was often plain old engineering mass.
Large software suites took years of work from huge teams. They needed specialists in UI, graphics, compilers, file formats, cloud sync, rendering, security, QA, accessibility, and platform support. Even copying the basics was a mountain of labor.
That’s why the ArtCraft example hits a nerve.
AI coding tools don’t remove every moat. They do weaken one of the oldest ones: the number of hours needed to write the first usable version.
A developer can now ask an AI model to help scaffold a Rust app, generate UI components, write file parsers, create test cases, explain compiler errors, refactor modules, draft documentation, and suggest architecture. The developer still has to judge, test, fix, and steer the work. But the loop is faster.
A task that once meant reading docs for half a day can become a conversation:
“Generate a basic image layer system.”
“Add undo and redo.”
“Write tests for layer ordering.”
“Convert this module to use a safer ownership model.”
“Find the bug in this rendering path.”
“Explain why this Rust lifetime error is happening.”
That doesn’t make the developer passive. It makes the developer more like a director, reviewer, and systems designer.
The bottleneck moves. It’s less about typing every line yourself. It’s more about knowing what to ask for, spotting bad output, and shaping the code into something that can survive real users.
Rust makes sense for this kind of project
The choice of Rust matters.
Rust has become popular for systems software because it gives developers strong control over performance while reducing whole classes of memory bugs. That combination fits desktop software well, especially apps that deal with images, video, documents, and large files.
Creative software can’t feel sluggish. If a user drags a slider, scrubs a timeline, resizes a canvas, or scrolls through hundreds of photos, the app needs to respond quickly. Rust is built for speed, but it also forces more safety into the development process than older low-level languages usually do.
That’s useful when AI writes or rewrites chunks of code.
AI-generated code can be impressive, but it can also be wrong in boring ways. It might mishandle errors. It might miss edge cases. It might produce code that compiles but falls apart under stress. A strict compiler helps catch some of that before users do.
Rust won’t save a bad architecture. It won’t make a half-built app production-ready. But it gives an AI-assisted developer a stronger fence around dangerous mistakes.
That’s one reason this pairing is interesting. AI speeds up generation. Rust pushes back with rules.
That tension can be healthy.
These clones are not real replacements yet
The fastest way to misunderstand this story is to treat ArtCraft like a finished Adobe or Microsoft killer.
It isn’t.
The video summary describes the apps as early alpha. That matters. Alpha software is usually for testing, experimentation, and gathering feedback. It may not have full feature coverage. It may not be stable. It may not handle every file format correctly. It may lack polish in the places that professionals notice within minutes.
A Photoshop alternative needs more than crop, brushes, and filters. It needs years of behavior that users already know. A Premiere alternative needs reliable exports, codec support, timeline performance, media management, and crash recovery. A Word alternative needs document fidelity, comments, styles, track changes, tables, printing, and all the weird formatting cases people have built into old files.
That last 20 percent is brutal. In mature software, it can feel more like the last 80 percent.
There’s also the legal and ethical line. Recreating core ideas and workflows is normal in software. Copying protected assets, branding, source code, icons, or proprietary designs is not. Open-source clones have to be careful here. The safest path is compatibility with user expectations without lifting someone else’s copyrighted work.
So the useful takeaway isn’t “cancel your subscriptions tomorrow.”
It’s this: the floor has risen.
A motivated developer with an AI coding assistant can now build the first serious version of a complex app much faster than before. That changes what small teams can attempt.
Why open source makes the story more interesting
ArtCraft being open source changes the meaning of the project.
Closed-source AI clones can appear, disappear, or turn into thin wrappers around someone else’s API. Open source gives people a chance to inspect the code, learn from the architecture, submit fixes, and fork the project if they want to take it in another direction.
That matters for software that aims at big incumbents.
Open-source projects rarely win by matching every feature on day one. They win when they become useful enough for a specific group, then improve in public. Blender is the obvious example in creative software. It took years of work from a large community, but it became a serious tool for 3D artists, studios, educators, and hobbyists. LibreOffice is another example in productivity software. It doesn’t erase Microsoft Office, but it gives many users a capable free option.
ArtCraft could follow that pattern, or it could stall. Early attention doesn’t guarantee long-term maintenance.
Open-source software needs more than code generation. It needs issue triage, documentation, releases, tests, contributor guidelines, packaging, user support, and boring maintenance. AI can help with those too, but someone still has to care enough to keep the project alive.
That’s where the “solo developer plus AI” story gets more grounded.
AI can help create a big burst of code. Community can help decide whether that burst becomes durable software.
The real threat to big software companies is not a perfect clone
Big Tech doesn’t need to panic because an alpha clone exists.
Adobe and Microsoft have deep advantages. Their products are mature, integrated, trusted, and used by enormous customer bases. They have cloud services, enterprise contracts, plug-in ecosystems, training materials, support teams, and decades of compatibility work.
But the risk for them is still real.
The danger isn’t that one open-source project instantly replaces Creative Cloud or Office. The danger is that many small teams can now attack small pieces of those suites at the same time.
Think about what happens when AI-assisted developers build:
A fast photo editor for basic social and ecommerce work.
A simple PDF editor that handles the five tasks most people need.
A spreadsheet tool for a niche industry.
A video editor built around short-form clips.
A vector app focused only on icons and web graphics.
A writing tool that handles clean documents without enterprise clutter.
Each one can be worse than the giant suite overall, yet better for one specific job.
That’s where pressure builds.
If customers only use 10 percent of a product, a smaller tool only has to nail that 10 percent. AI makes it cheaper to build those focused tools. Open source makes it easier for others to shape them.
Big companies can respond, of course. They already use AI in their own development workflows. They can add AI features to their products. They can cut friction, improve pricing, and make their tools easier to learn.
But they can’t rely on code volume as protection the way they once could.
Software engineering is becoming more about taste and judgment
This is the part that should make developers pay attention.
If AI can generate more code, code itself becomes less rare. Good judgment becomes more valuable.
That includes knowing:
What to build first.
Which features users actually need.
Where AI output is risky.
How to test generated code.
When to throw code away.
How to design simple systems that won’t collapse later.
The ArtCraft story points toward a practical future for engineering. Developers won’t just write functions. They’ll manage agents, review pull requests from AI, design architecture, create test paths, and spend more time deciding whether the output is any good.
That’s not a downgrade. It’s a different skill mix.
A weak developer with AI can create a mess faster. A strong developer with AI can explore more ideas, build larger prototypes, and ship useful tools with a smaller team.
That distinction matters.
The companies and developers who benefit most won’t be the ones who blindly generate the most code. They’ll be the ones who can turn AI-generated code into reliable software.
The takeaway
Brandon Thomas and ArtCraft don’t prove that one person can replace Adobe or Microsoft.
They prove something more interesting: one person with a strong AI model, a serious programming language, and enough persistence can now build the foundation of software that used to require a large team just to begin.
That changes the starting line.
The old moat around big software was made of time, code, people, and complexity. Some of that moat still stands. Trust, polish, compatibility, and support still take real work. But the wall around “can anyone even build this?” is getting lower.
That’s why this story matters.
Not because alpha clones are ready to take over. Because the next useful open-source tool, niche editor, or lightweight office app may come from one developer, a small team, or a community that would’ve never had the resources to try before.
Comments