From Coder to Conductor: Why 2026 is the Year of the Software Manager

Let this article serve as a timestamp for the moment we ascended to the next level of software engineering.

I wanted to share how my workflow has fundamentally shifted over the last few months. If I look back to the end of 2025, I was still a “manual” developer in many ways. I was writing roughly 30% to 40% of the code I committed. At that time, AI was primarily a high-end power tool, mostly used for autocomplete within the IDE. Anything remotely complex was a gamble; asking the leading models of that era to handle architectural shifts often resulted in a “context-collapsed” mess. It was good at the repetitive and the well-defined, but it wasn’t a partner.

The December Shift

Like many of my peers, I noticed a significant shift in December 2025. This was since been documented by “WTF happened in 2025?” Suddenly, the models (specifically the leap to Claude 4.5 Opus) didn’t just get faster; they got coherent.

My output flipped overnight. I went from writing 40% of my code to committing 95% AI-generated logic. I began using it for everything: writing documentation, performing rigorous PR reviews, and identifying the edge cases I used to miss. It felt like I had finally hired a world-class junior developer I could coach in real-time. This shift was partly due to the models hitting an exponential stride, but also because I finally became comfortable with the “vibe” of agentic tools.

The Watershed Moment: Conductor

That trajectory continued through January 2026, until a colleague introduced me to **Conductor **earlier in February. I cannot express enough how much this has impacted my reality. This has been my true watershed moment.

I’ve always maintained a healthy dose of skepticism toward AI, and I still do. But the productivity gains between “Chat-based AI” and “Context-driven development” via Conductor are eons apart. By moving project awareness out of ephemeral chat logs and into persistent, version-controlled artefacts (like plan.md and spec.md), the tool no longer forgets the blueprints.

For the first time, I don’t feel like I’m fighting an LLM to get it to understand my repo. Instead I no longer write code, I have become a Reviewer and Manager of a team of digital developers.

The Double-Edged Sword

However, we must acknowledge that this shift is currently a double-edged sword. While my personal workflow has reached new heights, the landscape is becoming increasingly uneven, and several systemic barriers are beginning to harden:

• The Entry-Level Collapse: We are seeing a historic 73% drop in entry-level tech positions. When one Senior with an agent can do the work of a five-person team, the traditional “junior” rungs of the career ladder are being cut away.
• The “AI Tax”: Staying competitive now requires a $100+/month subscription “tax” for Pro-tier models and agentic platforms. For students or developers in emerging economies, this creates a productivity gap dictated by bank accounts rather than brilliance.
• The Policy Handbrake: Many developers are trapped in “AI-resistant” corporations. Bound by legacy security policies or the strict new compliance hurdles of the 2026 EU AI Act, they are forced to work with one hand tied behind their backs.

The Productivity Paradox

It is difficult to reconcile this. We are witnessing **Jevons Paradox** in real-time: as the cost of producing code drops toward zero, the demand for complex, integrated software is reaching an all-time high. But the tragedy of this transition is that while there is more work to be done than ever before, the “entry requirements” have shifted almost overnight.

The future is undoubtedly bright for the capabilities of software, but as we celebrate our newfound powers, we must ensure these barriers today don’t become permanent tomorrow. As with all technological leaps, those who fail to adopt and adapt will ultimately be left behind, but let’s ensure the ability to do so isn’t outwith their control.

Vive la revolution 🤖 (but let’s make sure everyone has a seat at the table)