I Don't Vibe Code

4 hours ago

The author explains personal reasons for not adopting 'vibe coding' or LLM-driven development, despite current hype.
Cites being a cheapskate, avoiding recurring costs for AI tools, and preferring free tools like Emacs.
Notes that as an experienced developer, AI hype resembles past low/no-code tool promises and that LLMs mainly address 'accidental complexity,' not 'essential complexity' in design.
Argues that real-world complexity and abstractions require careful human judgment, which LLMs lack metacognition to question data or design choices.
Highlights a case where DOGE (an organization) misinterpreted SSA data due to ignoring context, similar to LLM limitations.
Values friction in coding as a learning tool and architectural signal, which LLMs bypass, potentially leading to poorly designed systems.
Criticizes LLM marketing for promoting solitary work over collaboration, potentially eroding team roles and product quality.
Expresses love for programming as creative expression and emphasizes ownership and responsibility, which LLMs cannot replicate.
Raises ethical concerns about LLMs, including misuse and consumer risk-shifting, and notes the unhappy, overworked culture among LLM enthusiasts.
Mentions minor irritants like the overly nice tone of AI chatbots and the value of unfinished projects for learning and brainstorming.

Hasty Briefsbeta