May 5, 2026
Cheap code, pricey chaos
Lessons for Agentic Coding: What should we do when code is cheap?
When writing software gets dirt cheap, the comments say the real bill comes later
TLDR: The article says artificial intelligence makes it cheap to write software, so people should experiment more and spend their time on the difficult parts. But the community's real message was harsher: fast code can become a maintenance nightmare, and some fear entry-level jobs are already paying the price.
A new essay on so-called "agentic coding" — basically using artificial intelligence tools to help write software — argues that when code becomes cheap, builders should experiment more, rebuild often, write better tests, and focus their human energy on the truly hard stuff. Nice, tidy, inspiring stuff. But in the comments? That calm advice got hit with a wave of anxiety, side-eye, and gallows humor.
The loudest reaction was a giant "not so fast". One commenter summed up the dread in one brutal line: if you race ahead with machine-made code, what happens when it becomes 10 to 100 times more expensive to fix later because nobody understands it? Another fan-favorite phrase, "free as in puppies," stole the show. Translation for non-programmers: sure, getting the code might feel free, but feeding it, cleaning up after it, and keeping it safe is absolutely not. That line became the thread's mascot for people warning that cheap software writing does not mean cheap engineering.
Then the mood turned darker. A commenter from India said beginner hiring is already getting crushed, with internships drying up and entry-level jobs becoming harder to land. That injected real-world stakes into what could have been a nerdy productivity debate. Meanwhile, another camp shrugged and said most of these "new" lessons are just old-school good management with shiny artificial intelligence branding. Even the jokes had bite: one person recommended a tool literally called get-shit-done, which somehow feels like the perfect mascot for this whole messy era.
Key Points
- •The article presents 10 lessons intended as durable guidance for agentic coding with AI tools.
- •It argues that when code is cheap, implementation and frequent rebuilding become effective ways to learn and refine product direction.
- •It recommends end-to-end tests, documented intent, and continuously updated specifications to support repeated reimplementation.
- •The article says the most valuable work shifts toward difficult areas such as design, performance, security, resilience, and architecture.
- •It concludes that while AI can reduce the cost of generating code, maintenance, support, and security costs remain significant.