The Evolution of DevOps — From Ghent 2009 to Agentic DevOps in 2026
Seventeen years, five distinct stages. What each one solved, what it left unsolved, and where we actually are now.
DevOps is not a single event. Since Patrick Debois ran the first DevOpsDays in Ghent in 2009, the field has gone through five meaningful phase transitions. Each new stage emerged to solve a new limit created by the previous one.
This post compresses those seventeen years. The category itself has been redefined repeatedly. Agentic DevOps — what we are doing now — is the answer to a new limit, not a marketing pivot.
Stage 1 (2009–2013) — Tearing down the wall between Dev and Ops
The original DevOps was a cultural movement. Development optimized for change; operations optimized for stability. The point was to merge those goals into one team's responsibility.
Tools were simple — Chef, Puppet, Vagrant. Infrastructure as Code found its first foothold.
What it left unsolved: defining infra in code is fine, but if the deploy itself isn't automated, you still ship by hand.
Stage 2 (2014–2017) — CI/CD becomes default
Jenkins, Travis CI, CircleCI went mainstream. Triggering a build on every commit stopped being aspirational and became baseline.
The seeds of GitOps were planted here — the proposition that the git repo is the single source of truth. Microservices also exploded the surface area CI/CD had to cover.
What it left unsolved: builds were automated, but deployment decisions and rollbacks were still humans clicking buttons.
Stage 3 (2018–2020) — Containers and Kubernetes
Docker and Kubernetes became the operational standard. "Cloud native" stopped being marketing and started meaning something specific.
SRE (Site Reliability Engineering) emerged as a distinct role. Google's SRE Book (2016) became the standard text. Observability settled into the logs/metrics/traces triad.
What it left unsolved: tooling expanded, but each org built its own unique combination, and onboarding new engineers turned into archaeology.
Stage 4 (2021–2024) — Platform Engineering
Companies realized that every team building its own DevOps stack doesn't scale. Internal Developer Platforms (IDPs) emerged.
The defining word of this stage is Platform Engineering. Backstage, Crossplane, Argo CD became central. DevSecOps stopped being a separate category and was absorbed as a platform-default responsibility.
What it left unsolved: platforms became rich enough that the platform itself turned into the bottleneck.
Stage 5 (2025– ) — Agentic DevOps
Where we are now. LLM-based agents perform operational actions under the same authority model as humans. Not just code assistance — they enter the action plane: deploy, rollback, incident response.
The two questions that define this stage:
- How do we define an agent's authority boundary?
- How do we ensure agent actions land in the same audit trail as a human's?
Without answers to both, agents stay tools inside the IDE. They never enter the operational plane.
Recap — five accumulated stages
| Stage | Period | Keyword | Solved | Left unsolved |
|---|---|---|---|---|
| 1 | 2009–2013 | Culture · IaC | Dev/Ops split | Manual deploys |
| 2 | 2014–2017 | CI/CD | Build automation | Manual deploy decisions |
| 3 | 2018–2020 | Containers · SRE | Operational standard | Tool sprawl |
| 4 | 2021–2024 | Platform Engineering | Tool unification | Platform complexity |
| 5 | 2025– | Agentic DevOps | ? | Authority · audit |
Each stage accumulates — it does not replace the previous one. Stage 5 doesn't re-solve stages 1–4; it adds a new plane on top.
FAQ
Q. People keep saying "DevOps is dead." True? A. The word is worn out as marketing. The movement has refreshed its meaning every stage. Not dead — evolving.
Q. SRE vs DevOps? A. SRE is Google's implementation methodology of DevOps principles. DevOps is the what; SRE is one how. → See SRE vs Platform Engineering vs DevOps.
Q. Will Agentic DevOps replace people? A. Short term, no — it redraws the authority boundary. Humans keep decisions and approvals; agents reduce wall-clock time on the execution plane. → Also: Why On-Prem AI DevOps Agents Are Actually Hard.
Related posts
When a Product Manager Ships Code, Who Owns the Outage?
In organizations where product managers write code with AI and several agents work at once, who checks what before a change reaches production? A role design for verification, approval, and recovery that is independent of the author, grounded in NIST SSDF, SLSA provenance, and the Google SRE postmortem culture.
John Baek
An Eclipse Plugin for the Agent Era — Task Context Was Already the Problem 20 Years Ago
The problem Mylyn set out to solve in the mid-2000s is the agent context problem. What changed is that the thing reading that context is no longer only a person.
Yeongsang Kim
A VS Code Extension for the Agent Era — What Developers Look At Now
Once writing code got cheap, a developer's time moved to judging and approving. Here is why those jobs cannot live outside the editor, and the choices behind the CollabOps VS Code extension.
Seungbaek Lee