2 months ago
- The author created a beeswarm chart showing AI model releases over time, with each mark representing a model and its horizontal position indicating the publication date.
- They discuss the initial excitement of AI-generated code feeling like magic, but overuse leads to producing low-quality work and overwhelming PR reviews, forcing a choice between being a bottleneck or relying on agents with half-hearted efforts.
- Reflecting on the past, the author notes that coding involved problem exploration, decision-making, and team collaboration, with code evolving from fuzzy ideas to clear implementations as a medium for thought.
- They acknowledge that while reverting to old ways isn't desirable, achieving a flow state requires new workflows and tooling, similar to how IDEs evolved over decades to support coding practices.
- The history of programming is traced from punch cards to assembly, compilers, scripting languages, and libraries, each step abstracting details to allow faster and more ambitious creation.
- Critiquing current AI use, the author argues that relying on prompts and raw code reviews leads to unreadable code and agent overuse, whereas modern AI models, rooted in language and translation research, can robustly translate code across mediums.
- They propose that AI agents could create custom whiteboards or playgrounds for exploration, starting with images or prototypes instead of code, to maintain intuition, mental mapping, and team alignment.
- The conclusion emphasizes that if AI is only used to edit code faster, it risks losing the thinking process; instead, tools should enable human-centric thinking to make coding feel grounded and engaging.