Technical debt can be an enticing choice for developers, suggesting significant opportunities that justify short-term sacrifices in code quality. Much like financial debt, its value hinges on the potential return. If the payoff is high enough, taking on technical debt might make sense. However, the cost of fixing this debt escalates rapidly. As developers accrue more debt, their established problem-solving methods falter, leading to an increasing number of bugs with each change. Eventually, the cost of addressing the debt can exceed the benefits of new opportunities. AI has disrupted the traditional approach to managing technical debt. The idea of agentic coding suggests that as AI models advance, paying down technical debt will become easier. Developers might rationally choose to delay repayment, anticipating that future improvements will make it cheaper to fix issues later. This mindset encourages rapid feature development, pushing the limits of current tools and resulting in massive codebases—some exceeding 40,000 source lines of code per day. However, this strategy is risky. The assumption of continuous growth in AI capabilities is unsustainable. At some point, the benefits of these models will plateau, leaving developers with a tangled mess of code that no model can adequately address. When this happens, the technical debt will become unmanageable, with millions of lines of code beyond any human or automated tool's comprehension. In essence, the technical debt taken on becomes subprime, and the reckoning will be hard to face.