Will AI Invent a New Programming Language?

My perspective on the question nobody’s asking. What if the answer is a design file.

I need to be upfront about something before we start.

I’m a designer. I’ve worked on apps serving over 1.9 billion active users at TikTok, 50 million active users at Grab. My coding knowledge stopped at JavaScript. And I’m writing an article about programming languages.

That alone should tell you how much has changed.

Two years ago, I wouldn’t have had an opinion on this. Now I’m deep in Figma MCP, Claude Code, and vibe coding workflows and the question of what language things are built in is suddenly very relevant to my work. Not because I’m writing code. But because the line between designing and coding is dissolving so fast that the distinction might not matter soon.

Whether that’s exciting or terrifying depends on the day.

The mess under the hood

Here’s something most people outside big tech don’t realise. The apps you use every day are not built in one language.

TikTok? Parts Kotlin and Swift for native UI. A custom cross-platform framework called Lynx that ByteDance built internally and open-sourced in 2025 because nothing on the market could handle their complexity. Grab serving 50 million users across Southeast Asia? Similar patchwork. React Native here, native code there, custom bridges connecting everything.

This is not an exception. This is the norm.

A single design system, the one we spend months perfecting in Figma has to translate into multiple codebases, multiple languages, multiple rendering engines. And the translation is lossy. Every time. “Should this spacing be 16px or 20px?” is a Slack message that has been sent a billion times across the industry, and it exists because design files and code are fundamentally different representations of the same thing.

So when people talk about “design to code” or “design in code,” this is the part they’re skipping. It’s not just a handoff problem. It’s a multi-language, multi-platform infrastructure problem that nobody has cleanly solved.

Which got me thinking. If AI is rewriting how we build software… will it also rewrite what we build software with?

Every era invents the language it needs

Quick history, because it matters for what comes next.

The web needed a scripting language for browsers. Brendan Eich built JavaScript in 10 days in 1995. Mobile needed languages that could talk to hardware. Apple created Swift. Google backed Kotlin. Facebook looked at all of this and said “what if we could use JavaScript to build native apps?” and React was born. Machine learning needed something accessible to researchers who aren’t systems programmers. Python became the default, not because it’s fast, but because the ecosystem made it frictionless.

Every time the paradigm shifted, a new language emerged. Facebook invented React. Vercel invented Next.js. The pattern is consistent.

So here’s the question. AI is the biggest paradigm shift since the internet. What language or what thing will it invent next?

The “language doesn’t matter” argument

The numbers are hard to argue with.

AI writes roughly 30% of Microsoft’s code and over a quarter of Google’s, according to their own CEOs. Y Combinator reported that 25% of startups in their Winter 2025 batch had codebases that were 95% AI-generated. Stack Overflow’s 2025 Developer Survey found 80% of developers using AI coding tools weekly.

Andrej Karpathy coined “vibe coding” in early 2025 the idea of describing what you want in plain English and letting AI handle the implementation. Collins Dictionary named it Word of the Year. By early 2026, Karpathy himself said vibe coding is already passé. The new paradigm is “agentic engineering” you’re not writing code at all, you’re supervising agents that write it for you.

Jensen Huang, NVIDIA’s CEO, put it even more bluntly: “It is our job to create computing technology such that nobody has to program and that the programming language is human.”

If AI can translate natural language into Python, Swift, Kotlin, or Javascript equally well, then the programming language becomes an implementation detail. Like assembly language. You don’t think about assembly when you write Python. Maybe soon, you won’t think about Python when you describe an app.

There are still new languages emerging like Mojo, built by Chris Lattner (the creator of Swift), reads like Python but performs like C++, heading to 1.0 in 2026. But the bigger story isn’t any single new language. It’s the possibility that language choice itself becomes invisible.

OK Sherizan. So what does this mean for designers?

This is where I have to be honest about two things that are both true at the same time.

1) The “design is dead” argument is not wrong.

I know designers don’t want to hear this. I don’t particularly want to write it. But let’s look at what’s actually happening.

PMs and designers at Meta have publicly said they now show working products instead of UI prototypes. Designers are shipping MVPs through AI coding tools before developers even get briefed. Figma’s stock dropped roughly 85% from its post-IPO peak. Wall Street is calling it the “SaaSpocalypse” a trillion-dollar selloff in software stocks driven by fears that AI is eating the workflows these companies were built on.

Shadcn happened. A component library so well-designed, well-documented and so easy to prompt with that people started building dashboards, landing pages, and full interfaces directly with shadcn components with no Figma file needed. You describe what you want, AI generates production-ready React with shadcn, Tailwind, and Next.js. It looks good. It works. No designer touched it. I’m a huge fan of v0 and what Vercel has done for vibecoding.

Then frontier models got scary good. You can one-shot a full interactive dashboard with Claude Opus 4.6 or Gemini Pro. Not a mockup. Not a prototype. A working component with real interactions, real state management, real functionality. The kind of thing that used to take a designer days in Figma and a developer weeks to implement done in literally minutes by someone who never opened a design tool.

There are reports that programmer employment in the US fell 27.5% between 2023 and 2025. The squeeze isn’t just on developers. Entry-level design work, the screen-by-screen production that used to be someone’s first job is exactly the kind of task AI automates first.

If you’re a designer and you’re not at least a little nervous, you’re not paying attention.

2) Something interesting is happening at the edges.

At the same time that people are declaring design dead, a new category of tools is emerging that suggests the opposite.

Pencil.dev calls itself “Design Mode for Cursor.” It’s a Figma-like canvas that lives inside your IDE. Your design files sit in your Git repo. You design visually, AI generates the underlying React. Or you prompt AI, and it designs visually on the canvas. The direction goes both ways.

Paper.design renders every element as real HTML and CSS. No conversion step. You design with real flexbox layouts, real CSS properties, real font rendering. Right-click any element → Copy as React. They raised $4.2M from Accel, and their thesis is interesting: the canvas of the future IS code, you just shouldn’t have to know that.

Figma launched “Code to Canvas” with Anthropic that generates an interface in Claude Code, say “Send this to Figma,” and it lands as fully editable design layers. Plus Figma Make, Code Connect linking components to production code, and MCP integrations letting AI agents read and write to Figma files directly.

These tools aren’t treating design as a step before code. They’re treating design as code. Or code as design. The distinction is breaking down.

And Figma’s own State of the Designer 2026 report shows something that complicates the “design is dead” narrative: 89% of designers say they’re working faster with AI. 91% say AI tools actually improve their designs. Designers leaning into AI are 25% more likely to say they’re satisfied at work.

So which is it? Is design dying? Or is it evolving into something more powerful?

I think the honest answer is: both. At the same time. For different people.

A “what-if” I can’t stop thinking about

I want to be careful here. This is not a prediction. This is a thought experiment. Because I think it’s worth exploring even if it turns out to be wrong.

What if the next programming language isn’t a language at all? What if it’s a design file?

A designer izaias described Figma as “logic wearing a canvas costume.” And he’s not wrong. Every design is JSON. Every component is a node. Auto layout is flexbox. Constraints are code. We’ve been programming all along, we just called it designing.

Paper.design’s canvas renders real HTML/CSS. Pencil.dev stores designs as JSON. Figma’s Code Connect links components to production code. Shadcn gives you a component system that AI can compose into anything.

So here’s the what-if.

What if a design system, not the Figma library we know today, but a next-generation, language-agnostic representation of an interface becomes the thing AI compiles from? You design the interface. You define the interactions. You set the tokens, the spacing, the hierarchy.

And AI handles everything underneath like choosing Swift for iOS, Kotlin for Android, React for web, C++ for the performance layer, building the bridges between all of them automatically.

That messy multi-language architecture at TikTok and Grab? It disappears. Not because someone invented a universal language. But because the design file became the universal specification, and AI figures out the compilation.

Design as the last programming language.

Now, hold on. I can already hear the objections. And they’re valid.

This assumes AI gets good enough at the translation. Right now, it isn’t. Vibe-coded apps have 1.7x more major issues than human-written code according to a CodeRabbit analysis. Security vulnerabilities are 2.7x higher. A Lovable analysis found 170 out of 1,645 apps had data exposure issues. The gap between “looks right” and “works at scale” is enormous.

It also assumes design can encode enough information to replace code. Today’s design files don’t capture business logic, state management, error handling or the thousand edge cases that make software actually work. A beautiful Figma file tells you nothing about what happens when the API fails at 3am.

And it assumes the industry will converge on a standard representation. Figma files are proprietary. Paper.design uses its own format. Pencil uses .pen files. We can’t even agree on design tokens, let alone a universal compilation target.

So yeah. Maybe this is naive. Maybe design files as programming languages is a fantasy that ignores how software actually works at scale.

But.

Two years ago, the idea that a designer could one-shot a working dashboard by describing it to an AI was also a fantasy. And here we are.

Where this actually leaves us

I don’t know if design is the last programming language. I don’t know if programming languages will become invisible. Nobody does. Anyone who tells you they know exactly how this plays out is selling you something.

What I do know is this.

The old workflow aka design in Figma, hand off to developers, argue about spacing in Slack, ship something that doesn’t match the mockup is dying. That part isn’t controversial.

What replaces it is up for grabs. Maybe it’s designers coding directly. Maybe it’s AI agents that make the design-to-code gap irrelevant. Maybe it’s new tools like Paper and Pencil that merge the two entirely. Maybe it’s something we haven’t seen yet.

For developers: if programming languages become an implementation detail handled by AI, then the value shifts to systems thinking, architecture, and knowing what to build, not which syntax to build it in. Karpathy’s “agentic engineering” isn’t just a rebrand. It’s a fundamentally different job description.

For designers: if AI can generate any interface from a description, then the differentiator isn’t your ability to push pixels. It’s your ability to understand users, define problems, make intentional decisions, and apply taste that AI can’t replicate. Paper.design’s founders put it well: AI will help turn design into code, but it’s the designer’s role to decide what to build, steer the AI, and make sure it’s cohesive.

The people who thrive in this next era won’t be defined by their tool like Figma, Cursor, Claude, whatever comes next. They’ll be defined by their judgment.

And honestly? As someone whose coding stopped at JavaScript but who’s now deep in Claude Code and Figma MCP workflows, I find that both terrifying and kind of hopeful.

I just don’t think we should pretend we know how this ends.

---

What’s your take? Is design evolving or dying? Is the programming language question even the right question? I’m genuinely working this out in public . Hit reply, I want to hear what you think.