“You should never build a CMS”

Git vs. CMS for content editing

  • Some argue Git-as-CMS is “hellish” for growing marketing/comms teams; non‑technical staff need WYSIWYG, not branches and rebases.
  • Others counter that many people can learn basic technical tools, and Git’s strong versioning is exactly what CMS UIs often weaken, leading to broken links, inconsistent assets, and siloed workflows.
  • A middle view: Git is fine as a backend if you wrap it in a friendly UI or simple scripts; forcing raw Git on marketers is unrealistic, but “powered by Git” doesn’t have to mean “use Git directly.”

AI agents vs. traditional CMS

  • One side reads the original AI-company post as: agents work better on code than through a CMS abstraction, so for some teams a CMS is now pure overhead.
  • Others clarify the original author explicitly said many teams still need a CMS; the claim is narrower: if agents can safely manipulate code/content, non‑technical users may not need a GUI for simple edits.
  • Some foresee “agent‑first” tooling where CMS content, docs, and tickets are all manipulated through APIs and MCP‑style servers rather than manual web UIs.

Complexity, scale, and custom builds

  • Many agree with the article’s core: simple homegrown CMSes inevitably accrete complexity, just like ad‑hoc build scripts that grow into build systems.
  • Several commenters describe successful lightweight setups (folders + Markdown/YAML, synced via Dropbox or similar) that beat generic CMSes for small, specific sites and were quick to build—especially now with AI-assisted coding.
  • Others stress these work only with a willing developer in the loop and narrow requirements; most orgs with non‑dev staff and richer workflows are better off with a mature, managed CMS.

WordPress and the CMS ecosystem

  • WordPress is cited as evidence that CMS problems are largely solved for common cases: drafts, approvals, scaling, caching, non‑tech editing, even headless use.
  • Critics respond that for complex ecommerce, large catalogs, heavy localization, or intricate data models, WordPress becomes a plugin‑laden, fragile stack; at that point either specialized SaaS (Shopify, etc.) or fully custom systems may be superior.
  • High‑end enterprise CMSs (Sitecore, AEM, etc.) are noted as serving a different tier than WordPress or static/Git setups.

Bias, marketing, and ethics

  • Both the AI‑company post and the CMS‑vendor response are widely seen as marketing pieces, though some readers still find them technically insightful.
  • There is significant criticism of the CMS vendor publicly naming the former customer and individual from the original story; some view this as discourteous or even a potential confidentiality/privacy issue, others as fair response to public criticism.

AI-written style and trust in content

  • Multiple commenters feel the CMS article “reads like LLM output” (short dramatic sentences, certain headline patterns); others strongly disagree and see it as obviously human.
  • The author insists they wrote it by hand, acknowledging their style may have been influenced by heavy AI usage.
  • This sparks a broader concern: informal “LLM radar” is fallible, and casual AI accusations risk becoming the new generic “shill” accusation in online debates.