DHH says one person can now build the 5 percent of Premiere or Photoshop they need on Linux

DHH thinks the missing Adobe app on Linux is no longer a wall. On the Lex Fridman Podcast, David Heinemeier Hansson said one person can now start rebuilding the Mac or Windows software keeping them from switching, even Premiere or Photoshop, as long as they begin with the 5 percent they actually use.

That sounds deranged if you picture a lone developer recreating every corner of Adobe Premiere Pro, color panels, export presets, enterprise licensing weirdness, and all the little buttons that exist solely to humble you at 1:13 a.m. But DHH isn’t really selling a one-man Adobe. He’s selling something sneakier and more believable: the personal clone, the artisanal knockoff, the “I only need this bit” app built by a human with taste and an agent with no lunch break.

100% one person can.

The org chart eats the robot

DHH’s best point is not about code. It’s about companies. He has spent his career preaching small teams, fewer meetings, and the moral superiority of not turning every product decision into a hostage negotiation. So of course his theory of AI agents is that they work best when they are not routed through six people with calendar holds and a shared Figma file called Final_Final_New.

As soon as you’re having human teams work together on something, the bottleneck is rarely implementation. It’s human bandwidth and communication.

This is the part big-company AI decks tend to sand down into mush. DHH is saying the 10x fantasy dies the moment the agent becomes another worker trapped inside the approval machine. If a product manager, a designer, a VP, and a CTO all need to bless the thing, the agent may code fast, but the humans still move at human speed, which is to say, slowly, defensively, and with comments in Google Docs.

to get that magical 10x, 100x, in a few rare cases a 1,000x productivity boost, you have to interact with the agents directly.

He also takes a swing at the familiar question, if AI coding is so powerful, why haven’t Photoshop, Premiere, Office, and every ancient corporate interface been reborn overnight? His answer: because old software companies are built for the old speed. Their advantage becomes their ballast. They know how to run giant processes, not how to let one opinionated builder sprint at the machine.

The 5 percent doctrine

The search-friendly version of the claim, “one person can build Photoshop for Linux,” is the spicy one. The useful version is smaller. DHH invokes the old Microsoft Office joke: everyone only uses 5 percent, but everyone uses a different 5 percent. AI agents make that joke operational. Build your 5 percent. Ignore the rest. Suddenly the impossible app becomes a weirdly manageable personal tool.

If there’s a piece of software you depended on that bound you to Windows, it is completely within reach for you personally to start rewriting it.

That’s the claim worth arguing about on Reddit, because it hits exactly where desktop Linux has always hurt. Linux runs the internet, the cloud, your phone in disguise, and probably something in your kitchen. It has not run the mainstream creative desktop, mostly because professionals need specific apps with specific muscle memory. DHH says agents change the shape of that problem. Not by persuading Adobe to care. By letting users route around Adobe.

His proof is modest but real. He said that after moving away from macOS, he missed the clean Markdown writing environment of iA Writer, tried Typora, and then had an agent help build the simpler thing he wanted. Not a platform. Not a suite. A writing app.

And in I think about 20 minutes it had the first version. I started using and it wasn’t quite right. Within 2 days I’d given up Typora and I wrote and have written all of my essays since that moment in iA Writer.

That example is also the limit of the argument. A Markdown editor is not Premiere Pro. Video editing drags in codecs, hardware acceleration, effects pipelines, media management, weird camera formats, and the cursed goblin kingdom of export settings. DHH’s claim is strongest when it’s about taste-driven personal software and weakest when it starts wearing Adobe’s jacket.

Still, the provocation is useful. If the agent era produces better software, it may not arrive first as a gleaming replacement for the bloated apps everyone hates. It may arrive as a thousand smaller escapes, each built by someone who finally got impatient enough to stop waiting for the product roadmap.