A few months ago, we were sketching on whiteboards and asking a simple question:
What would it look like if AI lived inside a team's workflow and actually built alongside them?
We kept coming back to the same idea. AI shouldn't feel like a separate tool you jump into or something you pass along at the end of a process. It should work as a genuine collaborator, one that understands your codebase, your components, your standards, and your intent.
As we wrap up 2025, the most energizing part of this job has been watching teams use Builder in real production environments. Startups. Enterprises. Agencies. Internal product teams. Real users. Real constraints. Real shipping.
What stood out this year was how naturally teams started weaving Builder into how they already work. The workflows are organic. They start where conversations start. They adapt to individual roles. And they turn the messy, nonlinear process of building software into something fluid and collaborative.
Here's what we saw most.
PMs take a Slack thread, a Jira ticket, a messy idea, and move straight into implementation.
"Could we test this?"
"What if we tried a simpler version?"
"Can we ship a v1 by Friday?"
Instead of weeks of translation across roles, Builder turns that intent into a deployable experiment fast. Tickets become code. Meeting notes become working UI and logic. The distance between ideation and execution shrinks dramatically.
Designers import Figma and see it become production-ready UI code that matches the details they care about: component structure, tokens, spacing rules, and naming conventions.
What they get is real, working code, built with the same care and consistency the team would normally invest over many hours or days.
Developers keep telling us the same thing: Builder gives them time back for the parts of engineering they actually want to do: architecture, systems, edge cases, performance, and reliability.
Builder handles the repetitive plumbing and the scaffolding. Teams move faster without the codebase turning into a mess.
This year we pushed hard on one idea: AI shouldn't just generate code. It should support the full loop, from idea to deployment.
- Planning
- Prototyping
- Architecture
- Implementation
- Refactoring
- QA & testing
- Deployment
- Iteration
The goal is one continuous flow, with context that carries across roles, tools, and time.
- Fusion 1.0: The first AI agent fluent in product, design, and code, connected to your existing codebase so it can reason holistically and not in silos.
- Slack and Jira integrations: Build from where conversations already happen, reducing friction and meeting people where they work.
- Design system intelligence: Fusion deeply indexes your components and documentation so it always designs and codes just like you.
- A new desktop app for Mac and Windows: Faster performance, smoother workflows, and a better day-to-day building experience.
- MCP servers integration: Connect Fusion to your database, APIs, and tools (Supabase, Netlify, Notion, etc.) through MCP or build custom MCP servers for proprietary tools.
- Privacy Mode with client-controlled encryption: Enterprise-grade control over data and model interaction.
- Custom Instructions: Guide Builder to produce code that matches your preferences, standards, and constraints.
- Google Cloud Marketplace availability: Making enterprise deployment dramatically easier.
- Starter templates (custom + standard): Bootstrap projects quickly while staying aligned with best practices.
- Firebase Studio integrations: A big step toward full-stack workflows with real backend logic.
- Native mobile apps (private beta): Bringing the web workflow to iOS and Android with your existing design systems and components.
- Plan Mode: A space to think before generating: acceptance criteria, architecture notes, user stories, flows, all co-created in minutes.
- AI model selection: Choose the model that fits the task, from fast iterations to deeper reasoning. Opus 4.5, Sonnet 4.5, Gemini 3.0 Pro, GPT-5.1, GPT-5.1 Codex, GPT-5 Mini and more
- Custom roles + ACL: Control what permissions a user has, what projects they can access, and what code files they can touch
2025 was the year Builder stopped feeling like a single "tool" and started becoming something deeper: an intelligent layer that adapts to how your team builds.
It learns your components.It learns your naming conventions.It learns your coding style.It learns how your team collaborates.
And with every correction, every merge, every iteration, it becomes more aligned with how you build.
Teams tell us they're moving faster without stepping on each other. That's what we're most proud of, speed without sacrificing harmony in the development process.
Every project you build teaches us what the future of software development actually looks like in practice. What we're seeing is a shift from task automation to real collaboration, teams building with AI, not just next to it.
Thank you for experimenting, pushing boundaries, and giving us direct feedback. You're not just using Builder. You're shaping what this becomes.
Here's to building the future together,
Steve
Builder.io visually edits code, uses your design system, and sends pull requests.
Builder.io visually edits code, uses your design system, and sends pull requests.
Connect a Repo
