Systems, Stories, and Skills: A 2025 Year in Review

Source: Dev.to

Introduction

After a four‑year hiatus, 2025 was the year I finally returned to consistent blogging.
It was a year defined by transition, both for the industry and for my own career. In late 2024, I stepped into the Enablement and Platform Engineering space, marking my first role outside of a security organization since graduating college. From this new vantage point I’ve watched the Generative‑AI landscape shift from simple chat interfaces to autonomous agents, standardized tools, and reusable skills.

Beyond the Hype: The Rise of Utility

When ChatGPT was initially released, I was not impressed. Every week there was another article about an LLM hallucinating. Without the ability to interact with external data sources or drive change in external systems, I saw limited value.

In 2025, the industry shifted to standardized tool‑calling via Model Context Protocol (MCP).

• The focus moved beyond single‑turn chat responses and smart code‑completion.
• Agentic loops—agents that can verify implementations with bash commands or control a browser—demonstrated the power of coding agents.
• Rather than being an “all‑knowing” LLM, the emphasis shifted to usefulness and the ability to use tools (via MCP).

From Experimentation to Execution

Claude Code emerged early this year as a practical coding agent, with a number of challengers following as the space matured. These agents helped me get back into personal projects after the same hiatus I mentioned above. They dramatically reduced the cost of experimentation and helped me avoid rabbit‑holes caused by endless trial‑and‑error.

One of those projects was Verifi, a tool that makes it easier for developers to manage certificates across programming‑language ecosystems. Instead of getting bogged down in syntax, I co‑developed a well‑defined plan with the agent, driving toward the outcomes and non‑functional requirements I cared about. Building Verifi made it clear that, as agent adoption grows, the ability to articulate and evolve a plan will matter more than the mechanics of implementation.

The New Frontier: Context Engineering

Early on, I assumed that giving an agent access to more MCP servers would make it more capable. Adding too many tools, however, degraded performance—reinforcing that agent failures are often the result of system design, not model limitations.

• Prompt Engineering works well for single‑turn prompts.
• The tension between prompt size and tool richness pushed Context Engineering to the forefront in 2025.

Context Engineering focuses on keeping the context window from being overwhelmed by tool definitions, system prompts, and custom instructions layered on top of the conversation.

How I changed my evaluation criteria

• Instead of judging agents by the best‑case output of a single interaction, I now assess whether their behavior remains understandable and predictable across a longer workflow.

Strategies that emerged

The Agent Skills standard (originally created by Anthropic) helped ensure consistency across coding agents so the right context could be pulled in at the right time. I created example skills, including one around dependency management. As I outline in the post, an agent can now load a dependency‑management skill only when a package file changes, rather than loading it upfront for every instruction.

Looking Ahead: Skill Management in 2026

Similar to this year’s discussions around MCP and scaling toolsets, I expect a parallel conversation around managing skills as their adoption grows. To cap off the year, I used downtime between the holidays to prototype a potential solution.

skset – “Skill Sets”

I built skset (short for Skill Sets) to explore what a skill‑management ecosystem might look like. The initial focus is the basic problem of skills scattered across different agent directories.

In 2026, it will be interesting to see if a standardized marketplace for sharing skills publicly emerges beyond the vendor‑specific ecosystems we see today. Discussions are already underway about how coding agents will fundamentally reshape software engineering and how that shift will impact engineers throughout their careers.

Final Thoughts

• 2025 delivered massive improvements to base‑model performance, but those capabilities are just the foundation.
• As we head into 2026, the “downstream” impact on automated code review, social coding, and quality governance at scale will become clearer.

For me, 2025 was the definitive turning point for AI. As I wrote in Cloudy With a Chance of Context, this era feels remarkably similar to the shift I experienced with cloud in 2018—when cloud computing moved from emerging tech to a general best practice, challenging governance teams to keep up.

2026 will be the year we see if these autonomous workflows can truly meet the high bar of production‑grade engineering. Based on what I’ve observed building skset and working with context management, the challenge won’t be any single capability but orchestrating skills, sub‑agents, MCP servers, and other capabilities together effectively. Teams will need to:

1. Solve discovery without overwhelming context budgets.
2. Configure logical groupings per project.
3. Understand how these features interact.

The journey has just begun—stay tuned.