Is the Agile Manifesto Dead in the AI Era? Insights from Martin Fowler and Kent Beck

Hello, I’m Tony Bai.

Twenty‑five years ago, 17 leading developers gathered in a Utah ski lodge and signed the Agile Manifesto, a document that would shape software engineering for the next two decades.

Among those pioneers were two figures who have guided generations of programmers: Martin Fowler, author of Refactoring , and Kent Beck, the father of Extreme Programming (XP) and a co‑author of the manifesto.

Fast forward to today, when generative AI is sweeping the globe and “instant AI‑generated iteration” seems to eclipse traditional agile cycles. The question arises: Is agile dead? What engineering philosophies remain?

History repeats: AI is just another microprocessor

Kent Beck opens calmly, recalling his childhood. He likens today’s AI to the advent of the Intel 4004 microprocessor, which turned massive, immovable computers into devices anyone could imagine using.

“Before microprocessors, computers were huge, immovable machines. When the Intel 4004 appeared, we realized ‘wait, this is also a computer!’ Suddenly the space of what we could imagine expanded infinitely.”

Beck argues that AI, like microprocessors, object‑orientation, and the Internet before it, is simply an “imagination amplifier.” He admits he is using AI for “wild, ambitious projects” in Rust, acknowledging many failures as part of exploration.

Martin Fowler adds a “second‑order thinking” perspective, expressing deep skepticism about new technologies while insisting that even his skepticism must itself be questioned.

“You must find a perfect balance between skepticism and curiosity. I’m extremely skeptical of blockchain, but that skepticism itself must be doubted.”

Fowler confesses he initially dismissed tools like GitHub Copilot as garbage until reading Simon Willison’s blog, which taught him that mastering a tool requires learning how to use it well—mirroring the early ridicule of object‑orientation.

Breaking the illusion: Agile’s enemy is never waterfall

When asked whether AI’s promise of “faster, better, cheaper” aligns with the original agile intent, Beck delivers a blunt observation:

“In practice, companies don’t actually want faster, better, cheaper. Misaligned incentives punish those who truly pursue efficiency.”

Fowler agrees, noting that unlike the early days when agile had to convince enterprises of its importance, today no company dares ignore AI’s significance—yet that confidence is the biggest trap.

He warns that the original “agile transformation” often devolved into a “formalism disaster,” spawning a massive “agile industrial complex.” The same script is now replaying with AI, as consulting firms peddle “AI transformation” remedies without real technical understanding.

“AI is becoming the new ‘snake oil.’”

(The note explains that “snake oil” refers to 19th‑century quack remedies that promised exaggerated benefits without scientific basis.)

Architect’s ultimate question: AI is destroying programmers’ “social”

Beck argues AI is “re‑soloing programming,” isolating developers behind AI agents.

“Extreme Programming created a safe social environment for inherently unsocial programmers through pair programming and intense discussion. Now I see developers saying, ‘I have six agents, so I’m a small‑team manager.’ No, you’re just using six tools.”

He laments that the communal, chaotic, human‑centric process that once liberated programmers is being replaced by solitary interaction with cold AI bots.

Fowler shares a similar concern, asking whether future teams will be “one‑pizza teams” (because agents don’t eat pizza) or “two‑pizza teams with doubled efficiency.” He bets on the latter, emphasizing a hybrid model of two humans plus N AI agents for pair programming, preserving human control over AI direction while retaining valuable interaction.

“When AI takes three minutes to return a result, we can discuss variable naming or architecture. When it returns in fifteen seconds, we lose that conversation time.”

Craftsman’s dusk: AI steals the joy of refactoring

Beck’s closing monologue expresses a craftsman’s sorrow: the pleasure of painstakingly transforming a “code mountain” into an elegant artifact is fading.

“The obsessive joy of turning a messy file into a work of art is disappearing.” “I can still understand the macro goal, but I must shift focus from enjoying ‘sculpting code’ to enjoying ‘understanding the domain,’ because we’ve lost the leverage in code sculpting.”

Fowler offers a more actionable suggestion, noting the Venn diagram of Developer Experience and Agent Experience forms a perfect circle: code that is friendly to agents is also friendly to humans.

“Well‑modularized, clearly interfaced, and thoroughly tested code is easier for AI to handle. Our decades‑old craftsmanship isn’t obsolete; it has simply shifted from guiding humans to guiding AI.”

Conclusion: Hold onto the unchanging rocks amid uncertain tides

The hour‑long dialogue provides no concrete prompts or model recommendations, but the two seasoned thinkers distill four enduring principles:

Maintain absolute skepticism, including skepticism of your own doubts.

Design minimal experiments and verify bold claims yourself.

Never abandon human interaction; it fuels creativity.

Write clearer, more modular, well‑tested code—for yourself and for future AI collaborators.

Beck adds a poignant final thought: perhaps we must forgo the pleasure of “sculpting code” and instead relish the joy of “understanding the world.”