Why Are We Still Talking About Containers Ten Years Later? Insights from Kelsey Hightower

Hello, I’m Tony Bai.

At ContainerDays Hamburg 2025, cloud‑native legend Kelsey Hightower, who announced his retirement last year, delivered a thought‑provoking keynote. Instead of joining the AI hype, he asked the audience a soul‑searching question about containers.

The "Tsunami" Cycle of Unfinished Projects

Kelsey recalled three waves he witnessed in his career: Linux replacing Unix, the rise of DevOps, and the Docker/Kubernetes container revolution.

He identified a relentless "tsunami cycle":

Hotspot eruption : a new technology (e.g., Docker) appears, VC money pours in, and everyone talks about it.

Frenzied chase : teams ship “good enough” releases, prioritising speed over completeness.

Abandoned unfinished work : before the technology stabilises, the next hype (e.g., AI) arrives, causing many engineers to jump ship and leaving half‑finished projects behind.

“We’re like kids chasing a rolling ball; even the goalkeeper abandons the goal, leaving it wide open.”

This explains why, ten years later, we’re still discussing containers: the original work was never fully completed, leaving behind complexity, incompatibilities, and fragmented “enterprise‑grade” distributions.

Apple’s “Unsexy” Work – The Future of Containers

Kelsey revealed that Apple is now integrating a native container runtime directly into macOS. This is not Docker Desktop or a VM‑in‑VM solution; it is OS‑level support provided by the open‑source project apple/container on GitHub.

He noted that Docker veteran Michael Crosby, now at Apple, is contributing to this effort.

Standardisation : container runtimes will become as ubiquitous as the TCP/IP stack, available on Linux, macOS, and Windows.

Invisibility : users will no longer need to install Docker or manage runtimes; containers will be as natural as water and electricity.

App‑store re‑architecture : future App Store distributions may consist of container images, solving dependency conflicts and sandboxing issues.

These developments represent the “final stage” of container technology for those who stayed at the “goal” instead of chasing AI.

On AI – Don’t Be a Blind Copier

Although a former Google engineer, Kelsey warned against the current LLM frenzy. He demonstrated an experiment asking a locally‑run LLM, “What version does FreeBSD Service Jails need?” The AI answered “FreeBSD 13 (nonsense)”, while the truth is “FreeBSD 15 (unreleased).”

His advice:

Don’t worship generation : don’t blindly use AI‑generated code, just as you wouldn’t copy Stack Overflow snippets without scrutiny.

Context is king : AI is a powerful search engine; provide the correct context to obtain useful answers.

Train yourself before training models : become a competent engineer first; only then can you judge whether an AI response is genius or garbage.

Final Counsel for Technologists

Career : view your career as a relay race—when you reach the peak, think about passing the baton, not clinging to the spot.

Open‑source : ignore commercial licence games; true open‑source spirit is about sharing and collaboration, not control.

Focus : emulate the German toolmaker Knipex—master one thing. The industry lacks craftsmen who can polish a technology until it becomes “boring” and “invisible”.

Conclusion

Kelsey’s talk serves as a sobering reminder to the restless tech community. He stresses that a technology’s real value lies not in its novelty but in its ability to solve problems and be delivered completely. In an era dominated by AI chatter, we should pay attention to the forgotten “goals” and finish the great engineering projects that truly matter.

Reference video: https://www.youtube.com/watch?v=x1t2GPChhX8