You're Not Taking On Enough Tech Debt | SingulariTea ☕

The cost of servicing technical debt is dropping fast because of LLMs, and that changes the entire calculus of when to take shortcuts. The author argues you should actually take more technical debt now and bet on future coding models to clean it up later. Think of it like finance: when interest rates are low, you borrow more. LLMs just crashed the interest rate on technical debt.

The piece extends the tech debt metaphor further than most. Not all debt is bad - taking on 20% interest to buy a machine that generates 900k profit in year one is smart business. The same logic applies to code shortcuts. If you can ship faster and grow revenue, throwing resources at cleanup later makes sense. The key insight: LLMs are changing what "later" costs. When a future AI can refactor your mess in minutes instead of weeks, the math on taking shortcuts completely changes.

Here's the uncomfortable part: you need to be a slightly worse coder now. All those maxims about clean code, helpful comments, and leaving code better than you found it? Still true, but less important. Most code going forward will be read by machines, not humans. The author calls this software's ZIRP era - when capital flies to stupid things because money is cheap. Except unlike economic cycles, this interest rate probably won't rise again. The cost of writing useful code keeps falling, which means the bar for what counts as "good enough" keeps dropping. Vibe coders might actually be calibrated better for this future than engineers trained to handcraft every routine.