Saying goodbye to Agile
Claims that “Agile is dead” in the age of AI-generated code have reignited long‑standing tensions between heavyweight process, light‑weight iteration, and up‑front specifications. Commenters contrast the original agile manifesto’s emphasis on small teams, feedback loops, and working software with Capital‑A Agile as practiced in many organizations, which they see as ceremony‑heavy, metrics‑driven, and often indistinguishable from fast waterfall. While large language models are pushing teams to write clearer specs for agents to consume, most argue this doesn’t replace iterative development so much as demand a better balance between specification, experimentation, and organizational reality.
Scope of “Agile” vs “agile”
- Many distinguish between:
- “agile” (lowercase): a loose ethos—short feedback loops, collaboration, adaptability.
- “Agile”/Scrum/SAFe: formalized processes with rituals, roles, and certifications.
- Several argue the manifesto is just values; the real problems come from rigid process cargo‑culting labeled as Agile.
- Others counter that, in practice, “Agile” now means those heavyweight processes, so retreating to the manifesto to deflect criticism is evasive.
Ceremonies, Metrics, and Dysfunction
- Common complaints: long “standups”, excessive meetings, planning poker, PI planning, ticket bureaucracy, and story‑point theater.
- Some describe Agile as a metric-production machine for management, or “waterfall done quickly” with sprints.
- Examples of gaming: doing work a sprint ahead to always “hit” estimates; inflating estimates to appear accurate.
- A recurring theme: processes imposed top‑down without team autonomy are demotivating and often ineffective.
LLMs, Specs, and “Spec‑Driven Development”
- Several note AI coding tools are pushing teams back toward clearer, more detailed specs or design docs.
- One camp claims this “exposes” Agile’s focus on coding speed as misguided and elevates specifications as the true bottleneck.
- Others respond that:
- Good specs were always the hard part; that’s why iterative/agile methods exist.
- Specs are often wrong or incomplete until users see working software, so iteration remains essential—even with LLMs.
- Hybrid views: richer AI‑assisted specs plus rapid agentic implementation cycles are framed as a new, highly iterative style that is still fundamentally agile.
Ideology, Cult Dynamics, and “You Did It Wrong”
- Many see a pattern: when Agile fails, defenders say it wasn’t “done right” or “enough,” likening this to religious or political dogma and sunk‑cost thinking.
- Others argue that most failures clearly violate core agile principles (e.g., no real collaboration, fixed feature roadmaps, no feedback), so “not doing it right” is often literally true.
- Several broaden this to a general criticism of project‑management “voodoo” and consultant‑driven process industries.
Context, Scale, and Alternatives
- Agile is reported to work well for:
- Small, competent, empowered teams with strong feedback loops.
- Less success is reported in:
- Large, management‑heavy organizations, or environments with hard external deadlines and rigid roadmaps.
- Alternatives and variants mentioned: Kanban‑style flow, “Flight” methodology, documentation‑driven development, and simply evolving team‑specific processes without labels.