I've been experimenting with the latest iterations of LLMs like Llama3, GPT-4-Turbo, and Gemini-1.5, and I'm impressed by their progress.
Initially, creating full programs from a prompt was unreliable, but these new models are starting to deliver.
For example, I used tulp (with gpt-4o and gemini-1.5) to write Arkanoid, Tetris, and Snake games in seconds, and the results were impressive. While these models already have knowledge of the tasks because they are well-known games, they have the ability to easily understand and add changes by prompt. The advancements feel quite remarkable.
What's your take on the current state and future potential of NO-CODE programming with LLMs?
TULP_OPENAI_API_KEY=YOUR_API_KEY tulp -x "A Snake game with Pygame. Press 'x' to start/restart, 'q' to exit. The snake starts small and grows when it eats. Food appears randomly over time. Red food is poisonous. Increase poison and speed with score. Make it colorful. The player has three lives. Include score and high score."
At the following link you can see more examples and some videos of the working games: https://www.linkedin.com/pulse/no-code-programming-federico-...