Google Cloud Next '26• Apr 23, 2026• 10:29ConferenceInterview

From idea to production: Building AI apps with GDE Tomek Porozynski

From Google Cloud Next '26 · 2026

Tomek Porozynski•Google Developer Expert (GDE); Cloud System Architect, PwC Polska

Executive Summary

Google Developer Expert Tomek Porozynski details his open-source project that uses Gemini's text-to-speech APIs to create multi-voice audiobooks, demonstrating a practical application of modern AI.
The discussion highlights specific AI development tools for different user levels: Google AI Studio for non-coders to rapidly build and deploy apps, and Antigravity IDE for experienced developers.
A key technique discussed is the use of "skills" to provide AI models with up-to-date information and documentation links, overcoming the challenge of rapidly evolving APIs.
The conversation emphasizes an iterative and collaborative development process, encouraging developers to brainstorm with AI agents, refine prompts, and use AI to troubleshoot imperfect initial results.

11 quotes

Concerns Raised

The rapid pace of AI innovation can lead to a 'fear of missing out' (FOMO) for developers.
Generative AI solutions often don't work perfectly on the first try and require iteration and refinement.

Opportunities Identified

Rapidly build and deploy full-stack AI applications from a simple prompt using Google AI Studio.
Automate application testing and troubleshooting using AI IDEs like Antigravity that can interact with a browser.
Create complex, high-value media like multi-voice audiobooks by orchestrating multiple AI APIs.
Democratize AI development, enabling individuals without coding skills to build and launch applications.

Key Themes

AI Development Tooling for Different Skill Levels

The speaker provides clear recommendations for getting started with AI development based on coding proficiency. Google AI Studio is highlighted as a powerful no-code/low-code entry point for rapid prototyping and deployment, while Antigravity is presented as an AI-native IDE for seasoned developers.

This provides a clear roadmap for individuals and teams to select the right tools, lowering the barrier to entry for AI application development and accelerating innovation for experienced practitioners.

Practical AI Application: Multi-Voice Audio Generation

The core of the discussion is a real-world project that converts text into a multi-voice audiobook. This involves complex steps like character identification, voice assignment, parallel processing of text-to-speech requests, and final audio assembly.

This project serves as a tangible example of how generative AI can be orchestrated to create sophisticated, high-value media content, moving beyond simple text generation.

Overcoming AI Model Limitations

A significant challenge in AI development is that models can have outdated knowledge of fast-changing APIs. The speaker explains how to use "skills" to provide the AI agent with links to current documentation, ensuring the generated code is accurate and functional.

This is a critical, practical technique for building reliable applications on top of rapidly evolving AI platforms, ensuring that development is not hampered by the model's last training date.

Iterative and Collaborative AI Development Workflow

The speaker advocates for a workflow that treats the AI as a collaborative partner. This involves an extensive planning and brainstorming phase with the AI before coding, as well as using the AI to help debug and refine its own imperfect outputs.

This insight reframes AI development from a simple command-and-response interaction to a continuous dialogue, leading to better-architected and more robust final applications.

Get started free

Topics

AI Development Generative AI Google Gemini Text-to-Speech AI Tooling Google AI Studio Antigravity IDE No-Code AI Developer Experience AI Agents Prompt Engineering Google Cloud Run Firebase Firestore Open Source AI Audiobook Generation

Processed Apr 28, 2026 yt-dlp + mlx-whisper + Gemini