January 26, 2026
From vibes to scribes
After two years of vibecoding, I'm back to writing by hand
Dev dumps AI vibes after finding "slop"; commenters split
TLDR: A developer says AI agents make neat snippets but chaotic projects, so they’re coding by hand again. Comments split: skeptics feel vindicated, a teacher warns AI stunts learning, and others insist better human steering fixes it—raising big questions about trust, skills, and how we build software.
A self-proclaimed “vibecoder” says AI agents write code that looks great line-by-line but turns into a messy novel when you read the whole chapter—so he’s back to hand-writing code. The crowd went wild. One fan joked he’s been vibecoding longer than vibecoding has existed, crowning him a trailblazer, while reluctant adopters cheered, “Finally, someone said it out loud.” The word “slop” became the day’s meme, with users posting imaginary “slop alerts” whenever agents got too confident.
Not everyone agreed. A CS teacher warned that because AI does easy tasks too well, new programmers skip learning the basics and never build real skills—cue academic panic. On the flip side, a pragmatist shot back, “That’s your job,” arguing humans should steer agents and update designs on the fly: tell the bot the new plan and make it obey. Another commenter pushed for guardrails and actually knowing your own codebase. The vibe vs scribe battle raged, complete with jokes about handing specs to a “mid-level engineer on vacation” and links to the dev’s X @atmoio. Verdict? The community is split between “ship by hand” purists and “manage the bot better” realists—and everyone agrees the slop is real.
Key Points
- •The author initially saw strong results from AI coding agents on simple and larger tasks.
- •Detailed, spec-driven prompts did not resolve issues because agents struggled to evolve designs over time.
- •Agent-generated code appeared plausible in isolation and in pull requests but lacked holistic structural integrity.
- •After reviewing entire codebases, the author found cumulative quality problems (“slop”) and misaligned decisions.
- •The author reverted to hand-coding, reporting better overall speed, accuracy, creativity, and reliability.