January 9, 2026
Vibes vs. bytes: choose chaos
AI Zealotry
Senior dev calls AI 'more fun'—comments cry junk, lost craft, and Big Tech power grabs
TLDR: A senior engineer says AI makes coding more enjoyable and efficient, advocating “vibe coding” with tools like Claude Code. The community clapped back: critics call AI output junk, fear automation erasing craft, and warn it boosts Big Tech power at society’s expense, making this shift both tempting and troubling.
A seasoned engineer posted AI Zealotry, insisting that using AI tools like Claude Code makes coding more fun: more thinking and writing, less “wrestling the computer.” He champions “vibe coding” (loosely exploring with AI), climbing the “abstraction ladder,” and using “Hooks” to stop mindlessly clicking “yes, allow.” The author admits downsides, but argues it’s like when we stopped writing assembly code by hand—trade some depth, gain a lot of speed. The comments? Absolute fireworks. One skeptic mocked the fun claim with a painter’s analogy: loving art but hating the brush. Another dropped a meme-level bullet list—“LLMs generate junk / LLMs generate a lot of junk”—calling out large language models (AI that generates text) for messy output. A third pushed back hard on the idea that zoomed-in coding skill is “obsolete,” pointing to ongoing mistakes and AI limits. The debate split between “AI as power tool for pros” and “AI as shortcut that erases craft.” One commenter warned this thinking leads to a future where writing and reading code is “entirely automated.” Then came the big-picture gut punch: AI dev might be easier, but it’s making people less free while supercharging Big Tech. Meanwhile, jokes flew about agents forgetting your CLAUDE.md like a gym routine and “pressing allow to your soul.” Spicy, chaotic, addictive.
Key Points
- •The author, a senior engineer, finds AI development more enjoyable and faster, enabling work in areas like frontend.
- •They acknowledge concerns: LLMs can generate junk, manual coding builds understanding, reviewing correctness is the bottleneck, and workflows can feel dehumanizing.
- •The transition to AI-assisted coding is compared to moving from hand-written assembly to using compilers like gcc, with trade-offs in understanding vs. productivity.
- •A key strategy is minimizing interruptions and automating simple tasks to climb the abstraction hierarchy and gain leverage.
- •Practical tip: use Hooks in Claude Code to enforce instructions (e.g., running tests with “uv run pytest tests”) since files like CLAUDE.md/AGENTS.md are often ignored.