December 27, 2025
Luck-Driven Dev vs RTFM
C –> Java != Java –> LLM
LLMs won’t rewrite coding rules—devs cry vibe chaos and dead docs
TLDR: The article says AI coding tools (LLMs) don’t replace programming languages; they still produce normal source code. Comments clash over 'vibe coding'—some say it kills docs, others call it a lottery, skeptics doubt live edits; it matters because workflows may stay the same even as AI speeds them up.
The article argues that AI coding tools—aka LLMs (large language models)—aren’t replacing programming languages the way C replaced assembly or Java replaced C. The “middle step” is still plain old source code, not English prompts. Cue the comments section going full reality TV. One camp says vibe coding (prompting the bot until it works) is already here, and it’s wrecking documentation. SadWebDeveloper declared the RTFM (“read the freaking manual”) era dead and blamed bots for copy-paste docs flooding the internet. Another camp shot back: specs (requirements) are too fuzzy, so “vibe coding” is basically buying a lottery ticket—cadamsdotcom roasted it as “luck driven development.”
Then came the live-coding drama. The author hints that interpreted languages (the kind you can run and tweak on the fly) may be the new hotness, letting you change running programs instantly. Some cheered, imagining “vibe coding to the max.” Others, like chuckledog, raised eyebrows: are we talking harmless config changes or full-on hot-swapping code mid-flight? The memes wrote themselves—“Dev Hunger Games,” “RTFM gang is over,” and “Ship it, we’ll refactor later.” The vibe: entertaining chaos meets cautious pragmatism, with developers split between hype, headache, and hopeful hackery.
Key Points
- •The article argues that past transitions between programming languages changed the intermediate product of software development (source code), altering tools and team practices.
- •It states that LLMs, unlike language transitions, do not change the intermediate artifact; developers still go from prompts to source code and then to binaries.
- •English prompts are not considered the intermediate product; source code in existing languages (e.g., Java, C, Rust, Python) remains central.
- •Until LLMs become fully autonomous, software architecture, storage, reviews, collaboration, and refactoring largely remain the same, though accelerated by LLM support.
- •The article suggests a potential shift toward dynamic interpreted languages to enable live, prompt-driven changes to running programs, reducing the run-refresh cycle.