Bluesky's Attie and Figma Make Just Brought Vibe Coding to Design and Social — Here's Why Non-Technical Builders Should Care
If you build software by describing what you want rather than writing it line by line, last week was quietly one of the most important weeks of the year. Two launches — one from Bluesky, one from Figma — pushed vibe coding out of the IDE and into the two environments non-technical builders spend most of their time in: design canvases and social products.
Neither announcement came with a flashy billion-dollar valuation. Neither needed to. Together, they signal something bigger than another model release: AI coding agents are moving upstream, closer to where ideas actually start.
Here's what happened, why it matters, and how to think about it if you're building software without a traditional engineering background.
Bluesky Launched Attie — An AI Assistant That Lets Anyone Vibe-Code a Social App
On March 28, Bluesky unveiled Attie, a standalone AI assistant built on top of the AT Protocol and powered by Anthropic's Claude. It was announced on stage at the Atmosphere conference by Jay Graber, Bluesky's former CEO and now chief innovation officer, and CTO Paul Frazee.
At launch, Attie does one very specific thing: it lets you build a custom social feed by describing what you want in plain English. Type "show me art posts from people I follow and similar creators" and Attie builds the feed. Type "tech news, but skip the crypto drama" and it shapes the algorithm around that constraint. No code. No dashboard. No algorithm-literacy prerequisite.
That alone would be notable. Custom social feeds have historically been a developer-only feature — you needed to understand APIs, data models, and ranking logic to build one. Attie collapses all of that into a conversation.
But the roadmap is where it gets interesting. According to Bluesky, the longer-term plan is to let users vibe-code their own social applications from scratch. Not just tweak a feed — build a new social product, end to end, by describing it.
That's a first. Until now, "build with AI" has mostly meant landing pages, CRUD apps, or internal tools. Attie is the first serious attempt to let non-technical builders ship social software — the single hardest product category in consumer tech — using nothing but natural language.
Why Attie Matters for Non-Technical Builders
Three things stand out.
First, it changes what "building an app" means. Most vibe coding tools assume your starting point is a blank project. Attie assumes your starting point is an existing protocol — AT Protocol — with millions of users, real content, and a live graph. You're not bootstrapping a product; you're shaping one. For a non-technical founder, that's an enormous head start. You don't need to solve the cold-start problem. You just need to describe your angle on it.
Second, Jay Graber's framing is unusually honest about who this is for. She told the Atmosphere audience that "AI should serve people, not platforms" — meaning the whole point is to give non-developers the power that used to belong to platform engineers. That's the vibe coding thesis, stated out loud, from a company that doesn't have to say it.
Third, it proves the category is expanding. A year ago, vibe coding meant Lovable, Bolt, and Replit — tools for building standalone web apps. Today it means a social network giving users a natural-language interface to build on top of its protocol. The surface area is growing, and it's growing toward the places where non-technical builders already have taste and instincts.
Figma Make Turned On AI Agents That Draw Directly on the Canvas
Two days after Attie, on April 2, Figma announced that AI agents can now design directly on the Figma canvas via the Figma MCP server. That sounds technical, so here's what it actually means if you're not a developer.
Until last week, AI design tools mostly worked in one of two modes. Either they generated images that you then had to manually recreate inside Figma, or they produced code that you couldn't easily edit in a visual tool. Either way, there was a translation tax between "AI generates something" and "you can work with it."
Figma Make closes that gap. Now an AI agent — running through the MCP protocol — can write directly into your Figma file, creating and modifying real components with your existing variables, tokens, and design system. When the agent builds a card, it's using your card component. When it picks a color, it's using a token from your palette. When it lays out a page, it's respecting your existing constraints.
Figma also introduced two concepts that matter more than they sound: skills and Make kits. Skills let you define how agents behave inside your file — what they can touch, what they can't, how they should reason about your design system. Make kits bring real components and data into the agent's context, so it starts from a grounded place rather than hallucinating from scratch.
Why Figma Make Matters for Non-Technical Builders
A huge share of non-technical builders — product managers, marketers, founders, designers who don't code — already live in Figma. It's where ideas get shape before they get shipped. By turning Figma into a place where AI agents can build, not just render, Figma is meeting non-technical builders exactly where they are.
It also changes the handoff. One of the most painful moments in shipping software without an engineering team is the gap between "here's a beautiful design" and "here's a working product." With agents that operate on real components and tokens, that handoff compresses. The same design system that drives your Figma file can drive the generated code downstream. You stop building one artifact in Figma and a completely separate one in your coding agent; you build one coherent thing across both.
This is the point where vibe coding starts to look less like a tool and more like a workflow. You describe what you want. An agent renders it in Figma using your real components. Another agent — maybe v0, maybe Voxel, maybe Cursor 3 — picks it up and ships it. You stay in the loop as the taste-maker and the decision-maker, not the implementer.
Taking a Positive Stance: Vibe Coding Just Got More Useful, Not Less Serious
Every few weeks, a new headline frames vibe coding as either a miracle or a disaster. The truth is less dramatic and more encouraging: it's becoming infrastructure.
Attie and Figma Make are evidence of that. They're not standalone "build a website with AI" demos. They're AI capabilities embedded inside tools you already use for thinking, designing, and shipping. That's what maturation looks like. A technology stops being a product category and starts being a feature of every product category.
For non-technical builders, this is the opposite of a threat. It means three things are happening at once:
The tools are getting better at respecting your work. Figma Make's agents write into your real design system instead of generating something generic. Attie builds on top of your actual social graph instead of a synthetic one. The days of "the AI rebuilt my project and lost all my context" are ending.
The tools are meeting you where you already work. You don't have to adopt a new IDE, a new deployment pipeline, or a new mental model. If you live in Figma, AI now lives there too. If you live on a social protocol, the same applies. The learning curve isn't "learn a new tool"; it's "use the tool you already know, but with a collaborator that writes code for you."
The tools are expanding the definition of what a non-technical builder can ship. A year ago, it was landing pages. Then it was CRUD apps. Then it was full-stack apps with databases. Now it's custom feeds, social products, and real design system work. Every quarter the ceiling moves up, and it moves up fastest for the people who were told for years they couldn't build software at all.
If you've been holding off on vibe coding because you thought it was a toy, last week is your signal. The toys are growing into tools. The tools are being plugged into the places you already live. And the builders who learn to use them now — before the category hardens — will have an advantage that compounds for years.
What to Do This Week
If you want to take last week's news and turn it into something you can actually use, three concrete next steps:
Get on the Attie waitlist. It's invite-only right now, but the public waitlist is open. Even if you don't have a social app idea, building a custom feed is a five-minute exercise that teaches you a lot about how agentic products feel when they're done well.
Try Figma Make on an existing file. If you already use Figma, open a real project — something with a proper design system — and ask an agent to generate a new page or component. Watch how it uses your tokens and components. That's the new baseline.
Pick a vibe coding agent and ship something small. Whether it's Voxel, v0, Cursor 3, Replit Agent 4, or Lovable, the point isn't which tool — it's the muscle memory. Describe a small thing. Let the agent build it. Review, iterate, ship. Repeat until it feels normal. That habit is the real skill, and it transfers to every tool that's coming next.
The category is moving fast, but the door is still wide open. Non-technical builders who show up with taste, judgment, and a willingness to describe what they want clearly are exactly the people these tools were made for.
Voxel is an autonomous AI coding agent built for non-technical founders, marketers, and product managers. Describe what you want to build — Voxel writes the code, tests it, and ships it.