Agile critique
The case against Agile as a methodology — anchored on two captures: Lewis Campbell’s “Agile in the dustbin of history” (clippings/lewiscampbell-tech-blog-260414.md) and the HN thread 28715218 (“Senior developers are leading the great resignation movement,” 335 pts, 2021 — clippings/hn-28715218.md).
The KRZ-endorsed anchor quote (ryandvm, HN 28717699, verbatim):
“Agile is like Communism. Sounds good in theory, but it’s never actually been used successfully. And every time someone tells a story about it failing the apologists inevitably reply with, ‘but that wasn’t real agile.’ I would suggest that if something is so hard to get right, then maybe it’s not such a great methodology after all.”
The “no true Scotsman” defense is the tell: a methodology that survives every failure by redefining itself out of the failure is unfalsifiable — and an unfalsifiable methodology isn’t engineering.
Campbell’s three arguments
- The Manifesto was content-free: “at best a sequence of vague platitudes (‘Customer collaboration over contract negotiation’), and at worst commercially unworkable (‘Welcome changing requirements, even late in development’).” Whenever questioned, apologists gatekeep “True Agile” — yet the source text “didn’t actually tell us much at all.”
- Agile was defined negatively — as not-Waterfall — but Waterfall was already known-dead in 1970: Royce’s own paper (the one that named the concept) recommended prototype-first, iterative refinement, and formal continuing customer involvement. Bell & Thayer 1976 (the paper that coined “Waterfall”) empirically showed requirements deficiencies surface only at design. Everything Agile claimed as innovation was 30 years old — “written the year after the moon landing” (the Apollo Guidance Computer shipped without Story Points).
- Spec-Driven Development is the reversal: Agile said “Working software over comprehensive documentation”; SDD says “Comprehensive documentation creates working software” — because LLMs (like good engineers) do badly with ambiguity. Royce again, 1970: “If the documentation is bad the design is bad. If the documentation does not yet exist there is as yet no design, only people thinking and talking… which is of some value, but not much.”
Where the vault already agrees
- First Principles Thinking — “everything that is not a law of nature is just a shared belief”; story points, sprints, and ceremonies are shared beliefs, and Campbell’s move is exactly decomposition to what’s irreducible (Royce’s actual findings).
- Golden repo / spec culture — specs as reviewable artifacts; the Agent skills thesis (procedural docs for agents) is SDD’s operational form.
- Agent knowledge systems / contextmaxxing — “documentation INSIDE the codebase, not attached”; Campbell’s SDD section is the same claim from the pre-LLM history side.
- The honest counterweight to keep: iterative/incremental delivery itself is not what died here — that part was real (Royce 1970, Boehm’s spiral). What Campbell puts in the dustbin is the brand, the coach industry, and the ceremonies that crowded out the substance. “Working software over comprehensive documentation” turned out to be false specifically at the moment software started writing software.
Related: First Principles Thinking · Golden repo · Agent skills · Agent knowledge systems · Working with legacy code (the fear/agile-adjacent org dynamics)