AI tools for human creativity

Key Concepts

• Vibe Coding: Using plain language to describe desired outcomes to an LLM, which then generates the necessary code.
• Google Flow: An AI-powered creative studio for image and video generation, featuring version history and custom tool-building capabilities.
• Project Genie: A world-modeling AI that transforms images or text into interactive, explorable 3D environments.
• Nano Banana: A high-fidelity image generation model used to steer and style outputs in Project Genie.
• Spaces: A feature in Flow Music that allows users to create, share, and iterate on custom music-coded apps and instruments.
• Generative Iteration: The process of refining AI outputs through multiple steps rather than relying on "one-shot" results.

---

1. The Creative Process and Fundamentals

The speakers emphasize that while AI tools are powerful, they do not replace the need for foundational design knowledge.

• The "Why" over the "What": Sanchit argues that asking "what prompt did you use?" is the wrong question. Success comes from understanding design fundamentals like Gestalt principles, composition, typography, and color theory.
• Additive Process: AI is not a replacement for traditional skills (drawing, 3D modeling, coding) but an additive layer. Professional experience allows creators to use precise technical vocabulary (e.g., "300mm lens," "ray-traced," "New Wave typography") in prompts to achieve specific, high-quality results.
• Iterative Craft: The speakers reject "one-shotting" (getting a result on the first try). True authorship in AI comes from a rigorous process of sketching, iterating, and refining—often going through dozens of variations to find the "one."

2. Vibe Coding: Democratizing Development

Callo and Khati demonstrate how non-coders and technologists alike use "vibe coding" to build bespoke tools.

• Methodology: By describing a desired interaction in plain language, the LLM writes the code. This allows designers to build their own custom software, such as "Super Looper" (an image processing chain) or "ASCII Film Studio."
• Real-World Application: Khati created "Machine Telephone," a game that passes inputs through a chain of different AI models (e.g., text to poem, poem to map location, map to song, song to color palette) to explore how context shifts across mediums.
• Responsibility: Callo warns against "slop"—creating content just for the sake of it. He argues that even with AI, creators must maintain high standards of taste and craft.

3. Music Creation: Personalization and Community

Kendall and Henry showcase how Flow Music enables both professional composers and casual users to create personalized audio experiences.

• Spaces: This feature allows users to build custom instruments or ambient soundscapes. Kendall demonstrated creating a "meditation space" by layering atmospheric filters and shimmering effects over a generated playlist.
• Practice Tools: Henry uses generative music to create specific accompaniments for drum exercises (e.g., triplet rolls), allowing him to practice with music that fits his exact technical requirements.
• Granular Synthesis: Henry demonstrated using AI to chop audio into tiny "grains" to create new, unique ambient textures, emphasizing that the more identity a creator puts into the process, the more expressive the result.

4. Project Genie: From Images to Interactive Worlds

Shosh and Sam explore the frontier of "world models," where static images become interactive game environments.

• World Remixing: Users can take a photo (e.g., a museum diorama, a toy car, or a pet) and turn it into an interactive 3D space.
• Physics and Biomechanics: The model understands how different objects move. For example, it can simulate the movement of a cat climbing furniture or a taxi driving on a road based on the physical attributes of the input image.
• Geo Genie: A new integration with Google Street View that allows users to select any real-world location as a starting frame and explore it as any character (e.g., a spaceship or a bird).

---

Notable Quotes

• "The question you should be asking is what was your thought process?" — Sanchit (on the importance of fundamentals over software).
• "The thing is, when stupid ideas work, they become genius ideas." — Khati.
• "I used to make things. Now, I make things that make things." — Sanchit (on the shift from being a designer to a tool-builder).
• "Just because we're using a model to code doesn't mean that we shouldn't be paying attention to every detail." — Callo.

Synthesis

The overarching theme of the presentation is that human imagination is the primary driver of AI technology. The speakers demonstrate that the most effective use of these tools is not to automate creativity, but to build personal, bespoke systems that reflect the creator's unique taste. By combining traditional design fundamentals with new generative capabilities, creators can move beyond standard UI and generic outputs to build richer, more meaningful, and highly interactive experiences.