CSS sucks because we don't bother learning it (2022)

CSS complexity and evolution

• Many argue CSS is inherently complex and has grown into a huge, constantly evolving surface area (multiple layout models: flow, float, table, flex, grid, etc.), making “knowing CSS” in full no longer realistic.
• Others counter that while new features are added, almost everything old still works; you can be productive with a subset (e.g., box model + flexbox or grid).
• Several comments note historical baggage: browser wars, early hacks (centering, table layouts, content-box vs border-box), and retrofitting new ideas for backward compatibility.

Declarative model, cascading, and “global state”

• Strong disagreement over mental models: some see CSS as “global mutable state” with class names acting like unscoped variables, causing brittleness and fear of collisions.
• Others insist that’s the wrong abstraction: CSS is a declarative rule system over a tree of boxes, more like a query/policy language than imperative code.
• Cascading and inheritance are described as both the core power and main source of confusion; debugging which selector “wins” is cognitively heavy.
• Newer features (@layer, @scope, custom properties scoping, Shadow DOM, Web Components) are seen as attempts to add proper modularity on top of originally global design.

Frameworks, Tailwind, and LLMs

• Tailwind and utility-first approaches are characterized by some as a symptom of not learning CSS; others say they pragmatically solve naming, scoping, and ergonomics problems.
• Component libraries (Bootstrap, MUI) are praised for making layout easier but criticized as leaky abstractions that still require real CSS knowledge.
• Multiple comments note that LLMs tend to emit “div soup” and non-semantic HTML/CSS, reflecting the poor patterns prevalent in scraped code, though some developers now lean on AI to avoid hand-writing CSS.

Ergonomics, quirks, and hacks

• Numerous specific pain points are cited: percentage padding tied to width, inline vs inline-block behavior, units that don’t map cleanly to pixels or physical sizes, historical difficulty of vertical centering, odd pseudo-class/element syntax, browser defaults like body margins, and recursive percentage sizing.
• Some argue these are necessary or historically contingent tradeoffs; others say they show CSS is conceptually flawed or “irreparably broken,” even if powerful.

Learning CSS: feasibility and attitudes

• Several recommend deep resources (notably a 1,000+ page “definitive guide” and specialized courses), emphasizing that treating CSS as a serious part of the stack transforms the experience.
• Others report years of effort, reading books and specs, and still conclude “CSS sucks” due to sheer volume of rules and edge cases.
• A recurring meta-point: many backend or full‑stack developers never fully commit to learning CSS, then blame the language; critics respond that even with commitment, it’s unusually full of gotchas compared with many programming languages.