Taste is the new programming language (w/ Astrocade CEO Amir Sadeghian) | AI Basics

Episode Summary

Executive Summary: The episode explores how Astrocade uses generative AI to let anyone create, customize, and publish interactive games and experiences from natural language, then refine them in an editor. Founder Amir Sa'Egian explains that AI reduces the traditional barriers of game development, but human taste, playtesting, and tuning remain essential. The conversation also covers creator monetization, personalized games, and the broader shift toward a new AI-driven creator economy.

Main Topics: Astrocade's AI-powered game creation platform (Priority: 5/5): Astrocade lets users describe a game or interactive experience in natural language, have AI generate it, and then publish it to a social feed of playable content. Lowering barriers to game development (Priority: 5/5): Amir argues that game creation has historically required coding, assets, tuning, and game-design expertise, and that generative AI removes much of that friction. Human-in-the-loop refinement and taste (Priority: 5/5): The episode emphasizes that AI can generate a first pass, but creators must iterate, playtest, and tune details like difficulty, pacing, and numbers for a fun result. Personalization and interactive experiences beyond games (Priority: 4/5): Astrocade supports more than games, including educational content, companions, birthday cards, and personalized experiences using uploaded images and custom assets. Creator economy and monetization (Priority: 5/5): The platform pays creators based on plays and has some users making thousands of dollars per month, positioning Astrocade as a possible new creator economy. Technical challenges and model improvement (Priority: 4/5): The team discusses training limitations, the need for deterministic controls in game systems, and plans to retrain models on their own interaction data to improve outputs. Hiring and product scaling (Priority: 3/5): Astrocade is hiring across product, recommendation systems, and AI post-training to better surface content and build stronger interactive-generation models.

Key Arguments: AI can dramatically reduce the skill and time required to make games, enabling non-experts to create interactive content. Generative AI is currently strongest at recreating familiar game patterns, but struggles more with novel mechanics and custom variations. Even with AI, good game design still depends on human taste, iteration, and playtesting because small parameter changes can significantly affect fun. The product works best as a hybrid workflow: natural-language prompting for creation plus an editor for precise tuning. Interactive content is broader than games and can include educational experiences, companions, and personalized creations like birthday cards. A large creator economy can emerge around AI-generated interactive experiences, similar to how YouTube enabled new kinds of creators. Training on proprietary interaction data may improve the models over time and create a self-evolving platform.

Data Points: Creators on platform: 100,000+ - Amir says creators from more than 100 countries are publishing content on Astrocade. Countries represented: 100+ - Creators are described as coming from more than 100 countries worldwide. Company founding year: 2022 - Amir says Astrocade was started in 2022 when generative AI became capable enough for the product vision. Office location at launch: in-laws' house - Amir recounts that the first office was in his in-laws' house. Watermelon juice experiment: small bottles consumed in 2 days - Used as an analogy for reducing friction and seeing demand appear once access becomes easy. Creator earnings: thousands of dollars per month - Amir says some creators make this amount on Astrocade through play-based payouts. Time to create a simple game: under 10 minutes - Jason notes he created a game very quickly while testing the product. Experience burden of traditional game development: 5-7 years of training - Jason contrasts the old path to making games with the new prompt-based workflow. Traditional production resources: five people, six months - Jason says a game like the demo would have taken a small team months to build previously.

Pivotal Quotes: "AI cannot make a full game. It can copy the games that they have been trained on." — Amir Sa'Egian: Amir explains current model limitations and why human creators still need to iterate and refine outputs. "Taste is the new programming language." — Amir Sa'Egian: He summarizes the product philosophy that creative judgment matters more than technical coding skills now. "Play is so universal, it's a universal human language." — Amir Sa'Egian: Used to explain why Astrocade aims to build a mass-market consumer product around interactive play.

Implications: AI is shifting game creation from a specialized technical craft to a more accessible, taste-driven creative process. For consumers, this means personalized interactive experiences; for creators, new monetization paths; for the industry, a likely wave of AI-assisted content and new tooling.