February 8, 2026
Hands-on > Hype-on?
I am happier writing code by hand
From dopamine hits to deep focus: why some coders are breaking up with AI
TLDR: An engineer says coding by hand beats chatbot “vibe coding” for focus, correctness, and happiness. The crowd splits: some want craft and depth, others love AI for freeing time and sanity, with jokes about slot-machine prompts and long-term mastery versus quick wins—making this a real debate about work and joy.
The author of the original essay says “vibe coding” with chatbots feels like an AI slot machine—fast dopamine, but deadened thinking—and they’re happier writing code by hand. Cue the comment wars. One camp cheers the craft: typing it yourself makes you truly understand the problem, catch mistakes, and stay in flow. Another camp counters: if AI buys back real life time, that’s the bigger win. As ramesh31 puts it, being unstuck and stepping away from the grind is the path to happiness.
The thread spit out analogies and memes like confetti. One user compared it to carpentry: machines can mass-produce furniture, but there’s joy in hand-carved work—even if it’s slower. acedTrex swore by long-term mastery, claiming hand-coding today pays off with faster shipping tomorrow. Meanwhile, falloutx joked about “agentic programmers begging for tokens on Kickstarter,” roasting the gambling-esque vibes of endless prompting. Others warned that big auto-generated code dumps still need review, leaving you the bottleneck anyway. The hottest take? Agents give breadth, humans give depth—and depth might be what businesses actually value. Whether you’re a neovim monk or a prompt cowboy, this fight isn’t about lines of code, it’s about joy, control, and responsibility—and the crowd is loudly split.
Key Points
- •The author reports repeatedly trying and then uninstalling claude‑code after a few weeks due to reduced motivation and engagement.
- •They state that writing code by hand helps internalize problem context and better assess correctness.
- •They describe “vibe coding” with LLMs as addictive and prone to passive acceptance of changes, even for trivial tasks.
- •Large, quickly generated pull requests still require review by the responsible developer, maintaining a bottleneck.
- •They propose a controlled use of Claude: manually provide context, restrict changes to limited scope, keep diffs small, and use it for tests or contained edits.