Why I Left to Build

I burned out. I kept functioning but the joy leaked out. Perfectionism and hustle culture optimized the curiosity out of me. Learning became a chore measured in career relevance. Building became a checklist.

Anxiety was driving most of my decisions. Not interest. Not conviction. Fear of falling behind, missing the window, not being enough. That is not a sustainable engine.

So I stopped.

The Break

I left big tech. Not because it was bad. The trade-off just stopped working. My day-to-day served someone else’s vision. The domains I actually care about were adjacent to my work, not central to it. I had scale but not sovereignty.

I wanted to build things I believed in. Understand systems from first principles rather than consuming them as abstractions. Have my work compound toward something I could see.

I used to love this. The elegance of a well-designed system. The satisfaction of making something work. Somewhere along the way that got buried under performance reviews and sprint velocity. Part of why I left is to find that again.

The Problem

AI tooling is transforming software development. But there is a critical gap no one is solving well.

You cannot review what these models produce with any certainty.

The volume and velocity of LLM-generated changes exceed human capacity to verify. Code review breaks down when a model produces hundreds of lines in seconds. The asymmetry is untenable.

Current approaches treat AI as a faster typist. They bolt generation capabilities onto existing workflows and hope the old verification methods scale. They do not.

The Work

I am building something different. An agentic SDLC where humans remain meaningfully in the loop.

Executable specifications. Not documentation that drifts from implementation, but specifications that are the implementation. Architecture as code that enforces its own constraints. Tooling that maintains security invariants automatically rather than relying on human vigilance at review time.

The goal is not to remove developers. It is to change what developers do. Less time verifying that code matches intent. More time defining intent clearly. Less fatigue reviewing generated output. More confidence that output conforms to declared constraints.

This is the work I am doing now. Making AI tooling deterministic enough to trust. Building the layer between human intent and machine execution that makes oversight possible at scale.

This Space

A paradigm shift is forming. The way software gets built is changing faster than our processes can adapt. Most of the discourse is noise. Hype cycles, fear, predictions that contradict each other weekly.

I am not interested in predicting the future. I am interested in building toward one that works.

This site is where I think through the transition. The posts will reflect whatever I am working on. Technical problems, architectural decisions, things I get wrong and have to revisit.

If any of that is useful to you, welcome.