We explored why 2025 was a turning point and how modern tools unlocked AI’s real potential. The next challenge lies in managing the pace of change these tools introduce.
From Long Cycles to Continuous Evolution
For decades, development environments changed slowly. Many developers used the same tools for 10, 15, or even 20 years.
At Consultingwerk, this meant nearly two decades shaped first by the OpenEdge UIB and later the AppBuilder, followed by almost another 20 years dominated by Eclipse-based tools such as OpenEdge Architect and Progress Developer Studio.
This long-term stability came to an abrupt end with the rise of AI, which fundamentally disrupted these established development habits.
Today, IDEs, AI agents, and underlying models evolve continuously:
- New large language models are introduced every few weeks
- Existing tools gain new capabilities such as parallel agents or advanced planning
- Updates are deployed automatically and frequently
This dynamic forces teams to continuously evaluate tools, models, and workflows.
At Consultingwerk, this shift became very tangible with our gradual move toward VS Code–based environments.
While Eclipse and Progress Developer Studio had shaped nearly two decades of daily development work, they increasingly appeared as limiting factors in an AI-driven world. Many modern AI extensions either do not exist for Eclipse or require versions that are not supported by Progress.
Windsurf - originally introduced to us as Codeium - marked a decisive step.
Built on top of VS Code, it combined a familiar editor experience with deeply integrated AI capabilities. Over the course of 2024 and 2025, Windsurf evolved rapidly:
- from advanced code completion
- to proactive refactoring suggestions
- to support for parallel AI agents
- and continuous model updates delivered in very short release cycles
Key shift: The IDE itself is no longer a static tool, but an evolving AI platform.
A Growing Ecosystem of Tools
Windsurf regularly introduced new features - sometimes major, sometimes incremental - and occasionally also surfaced challenges such as increased memory or CPU consumption, which were usually addressed just as quickly in subsequent releases.
At the same time, the ecosystem around VS Code continued to expand.
Alternatives such as Cursor, GitHub Copilot, Google’s Antigravity (based on Windsurf and Google Gemini), and third-party extensions like Kilo-Code offered different trade-offs in terms of speed, planning depth, and cost.
Beyond IDEs:
- Claude Code enabled CLI-based, prompt-driven refactoring
- Devin provided browser-based environments capable of analyzing entire codebases and answering architectural questions in natural language
Rather than committing to a single tool or model, we learned to treat this landscape as a toolbox - selecting the right combination of IDE, agent, and LLM depending on the task at hand.
Choosing the Right Model for the Job
Not all LLMs are equal.
Some excel at fast code generation, others at deep reasoning, refactoring, or test creation. Increasingly, developers deliberately choose different models for different tasks - balancing cost, speed, and quality.
Professional shift: Understanding AI behavior becomes part of software craftsmanship.
Why This Matters
Organizations that embrace change as a constant - rather than resisting it - gain a decisive advantage. Flexibility becomes a core competency.
Outlook to Part 4
As tools and workflows evolve, so does the role of the developer. In the next part, we examine how AI reshapes responsibilities, skills, and expectations.