Last CallLast Call Pool
Product & Management Stage

Jak dowozić z AI?

September 18, 2025 3:50 PM
Long Lecture
🇵🇱 Polish
Wisła

About

Zobacz, jak możesz wykorzystać AI jako wsparcie produktywności i dowozić projekty do końca, z pomocą nowoczesnych narzędzi. Skupimy się na praktycznych zastosowaniach nowych narzędzi, które pozwolą Ci tworzyć MVP aplikacji i projektować z większą swobodą. Dowiesz się także, jak stworzyć armię asystentów projektanta, którzy będą pracować z Tobą ramię w ramię.

‍

Talk by

Photo of Grzegorz Róg, CEO at easy_
Grzegorz Róg
CEO at easy_

Watch the full talk

Watch this WaysConf session, then continue with related talks or explore the current programme.

Explore the 2026 agendaMore WaysConf recordings
From the recording

Talk in brief

At WaysConf 2025, Grzegorz Róg, CEO of easy_, shares his practical workflow for vibe coding and delivering digital products using artificial intelligence. Moving past the traditional design-only role, Róg advocates for the vibe designer who designs, codes, and deploys side-by-side with AI. He demonstrates how to use Alice, a native desktop assistant app he co-developed, alongside tools like Cursor, Figma MCP, Vercel v0, Google AI Studio, and Claude Code. Róg provides step-by-step examples of generating design specs, prototyping UI components, converting web apps into native iOS apps, and automating project management tasks using Model Context Protocol integrations with Linear. Finally, he shares strategies for managing AI subscription costs through API-based pay-as-you-go models and building an intuitive daily habit of AI collaboration.

Key takeaways

  1. 01

    The Rise of the Vibe Designer

    Modern designers must transition from static design to active building, using AI to rapidly iterate, write code, and deploy functional prototypes.

    Watch from 1:30
  2. 02

    Leveraging Model Context Protocol (MCP)

    MCP allows AI assistants to directly interact with external tools like Figma and Linear, enabling seamless code generation from design frames and real-time task management.

    Watch from 7:32
  3. 03

    Optimizing Context with Modular Assistants

    Instead of dumping all files into a single massive AI context window, users should employ specialized, lightweight assistants for distinct tasks like design, marketing, or coding.

    Watch from 14:33
  4. 04

    Agentic Deployment with Claude Code

    Tools like Claude Code can run locally in a project directory, automatically handle environment setups, authorize hosting platforms like Vercel, and deploy applications directly from the terminal.

    Watch from 19:48
  5. 05

    Cost-Efficient AI via API Keys

    To avoid subscription fatigue from dozens of individual AI tools, users can connect their own API keys to a single interface like Alice, paying only for actual token consumption across multiple LLMs.

    Watch from 36:52
Read edited transcript highlights

These concise notes were edited from automatic captions and checked against the talk structure. They are not a verbatim transcript.

The Shift from Traditional Designer to Vibe Designer

The landscape of product design is shifting from isolated creative work to rapid, AI-assisted building. Traditional designers spend hours searching for inspiration and focusing solely on visual layouts. In contrast, the modern vibe designer works hand-in-hand with AI, allowing inspiration to flow naturally through rapid iterations. This new breed of designer does not just hand off static mockups; they actively write code, prototype interactive components, and deploy functional applications, significantly accelerating the entire product development lifecycle.

Watch from 1:30

Streamlining Code Generation with Figma MCP

Model Context Protocol (MCP) is a game-changer for bridging the gap between design and development. By connecting an editor like Cursor to Figma via an MCP server, the AI gains direct context of the selected design frames. When you ask the AI to code a component, it analyzes the auto-layout, margins, and visual hierarchy directly from Figma. By referencing a local Markdown file containing your team's specific design system guidelines, the AI outputs highly accurate, production-ready React or Vue code that perfectly respects your engineering standards.

Watch from 7:19

The Importance of Small, Specialized Contexts

A common mistake when working with LLMs is dragging every company document into a single chat window, expecting the AI to understand everything. In reality, large context windows often lead to hallucinations and diluted focus. A much better approach is to maintain small, highly specialized AI assistants. For example, you can have one assistant trained on design system guidelines and another on SEO and marketing. By summoning these specific assistants into a conversation only when needed, you keep the active context clean and get highly accurate, predictable results.

Watch from 13:34

Agentic Workflows with Claude Code

Moving from a local prototype to a live, deployed application is incredibly simple with agentic tools like Claude Code. By running a simple command in your terminal, you can run Claude directly inside your project directory. The AI agent understands the entire codebase, creates files, and executes terminal commands. If you ask it to deploy the app to Vercel, it will automatically detect if you are logged in, provide an authorization link if needed, package the files, and deploy the application to a live URL without you ever needing to write a line of deployment configuration.

Watch from 19:48

Porting Web Applications to Native iOS

Once you have a working web application, converting it into a native mobile app no longer requires hiring a dedicated mobile development team. By utilizing Git version control, you can create a new branch specifically for mobile development. Using terminal-based AI agents, you can prompt the system to refactor your web React code into a native iOS application. The AI handles the complex translation, allowing you to open the project directly in Xcode and run a fully functional native app in the iOS simulator with minimal manual intervention.

Watch from 23:51

Managing AI Costs with Pay-As-You-Go APIs

Many professionals suffer from subscription fatigue, paying for dozens of different AI tools every month. A highly effective way to manage these costs is to bypass consumer subscriptions and use developer API keys instead. By integrating your API keys into a unified desktop interface like Alice, you only pay for the exact number of tokens you consume. This allows you to seamlessly switch between top-tier models from OpenAI, Anthropic, and Google within a single conversation, often costing only a fraction of the price of multiple flat-rate monthly subscriptions.

Watch from 36:52
Explore WaysConf 2026

View the agenda

Get tickets