There will be a lot of throw aways but it's okay

AI is moving fast—faster than most of us can truly process. One week, a new model dominates the conversation; the next, an entire framework for agentic workflows makes the previous one feel like a relic. In this whirlwind, it’s easy to feel a sense of hesitation. Some people are waiting for the "perfect" moment to jump in, thinking, "I’ll wait until the tools are better," like someone holding off on buying a TV because next year’s model will have more pixels.

But there’s a fundamental shift happening in software engineering that makes this waiting game unnecessary. Software engineering is moving toward full automation, and with it, the concept of "sunken cost" is effectively hitting zero.

In the old world, code was an investment. You spent months architecting a system, documenting every line, and optimizing performance. If the paradigm changed a month later, it was a tragedy. That code represented thousands of hours of human life.

Enter vibe coding. When your AI agent is the one doing the heavy lifting—writing the boilerplate, refactoring the logic, and implementing the features based on your "vibe" and intent—the cost of throwing it all away vanishes. If a better tool or a more elegant architectural pattern emerges next month, you don't have to mourn the lost hours. You just prompt your agent to adapt.

It’s okay to build things that might be "useless" in six months. Those "throw-away" projects are the training ground where you learn to stay synchronized with the AI. You keep vibe coding away your next big ideas, and if the paradigm shifts, it’s absolutely fine. After all, it was your little AI agent coder who wrote the whole thing anyway. The real value isn't in the lines of code—it's in your ability to direct the agent that wrote them.

Note: This post was generated by Antigravity based on my obsidian note.