February 8, 2026
Bots vs brains: choose your fighter
Stop Generating, Start Thinking
Dev calls AI code “fast fashion” — comments explode with memes, clapbacks, and panic
TLDR: An experienced developer blasts AI‑generated code as low‑quality “fast fashion” and warns it’s wasteful and opaque. Commenters split: some defend AI with task‑based agents, others cite shaky results and security flubs, turning the thread into a battle over whether to think first or outsource to bots.
A veteran engineer says it out loud: stop “prompting” bots and start thinking. They compare large language model (LLM) code — the stuff chatbots spit out — to fast fashion: quick, cheap, and full of holes. Cue the crowd. Some cheered the takedown, citing energy waste and the “shrimp Jesus” meme as proof we’re using planet-sized power to make silly pictures (KnowYourMeme). Others argued the future is agents, not prompts. One user bragged they no longer “talk” to AI — they assign tasks and let multiple bots work like interns, messy but useful.
Then it got spicy. A blunt commenter roasted devs for craving instructions over thought: “Why think when you can be told what to do?” Meanwhile, a practical voice pointed to headlines about Microsoft dialing back Copilot sales targets and a “vibe coded” app with security holes, shouting: AI isn’t “there yet.” The vibe: equal parts hype and hangover. Amid the chaos, someone yelled the site’s text was unreadable, proving that accessibility drama never sleeps.
The author’s main beef? Non‑deterministic, opaque outputs — unlike reliable machinery. Fans of AI pushed back: use it like teammates, not magic. The thread split between “write real code” purists and agent wranglers, with Copilot (GitHub), Claude (Anthropic), and SHEIN‑style metaphors getting the tabloid treatment.
Key Points
- •The author uses AI tools like Copilot and Claude mainly for autocomplete and debugging but finds them unreliable for complex tasks.
- •Prompt and context management are described as time-consuming compared to writing code directly.
- •The article warns against widespread reliance on generated code due to quality and maintainability concerns.
- •Parallels are drawn between mechanization’s mass production and AI’s scale, highlighting environmental costs from data center energy use.
- •A distinction is made between deterministic mechanized processes (e.g., codemods) and non-deterministic, opaque LLM outputs.