One-Day Sprints

I used to manage software development teams. A sprint was two weeks. Five engineers. Maybe seven features if we were lucky. The coordination overhead was enormous — standups, code reviews, merge conflicts, context switching.
Now I run sprints in a day. By myself.
I’m building a voice pipeline — speech-to-text, language model reasoning, text-to-speech. Real-time, low-latency, running on my own hardware. The kind of project that would have taken a team months to prototype.
Here’s what changed: I work with models from three different companies simultaneously. Claude for architecture and complex reasoning. GPT for certain code patterns. Gemini for edge cases and validation. Each model has different strengths — trained on different data, optimized for different tasks. I treat them like a technical team with diverse expertise.
The collaboration isn’t sequential. It’s iterative. One model drafts, another critiques, a third refines. I’m the product manager and the integrator. The models are the team.
This isn’t vibe coding — throwing prompts at an AI and hoping for the best. It’s disciplined. I still write architecture docs. I still define acceptance criteria. I still review every line. But the cycle time collapsed from weeks to hours.
Twenty-five years of managing engineers taught me how to break down problems, sequence work, and integrate components. Those skills didn’t become obsolete. They became leverage.
The bottleneck isn’t coding anymore. It’s knowing what to build.