
Now building what developers will run on.
Software · Systems · Agents
What I Believe
Software rots. Teams can't keep up with framework upgrades, so they wait until EOL and then it's a scramble. Code reviews pile up because nobody has time to understand what changed. Cloud bills grow quietly because you're shipping too fast to check. The business thinks software is a non-degrading asset. Engineers know better.
The best tools I've built are mostly good engineering. Graph analysis, AST transforms, dependency tracing. AI shows up where it counts: eliminating the manual grind, generating explanations from real context, catching what humans skip when they're moving fast. Engineering does the work. AI removes the busywork.
This isn't speculation. I'm building these tools with the same rigor I brought to systems that move real money and serve billions of users.
What I'm Building
AI lets us move faster than ever. That speed creates its own problems. I'm building tools for both sides of that equation.
Codebases age. Frameworks move on. Migrataur handles migration at scale. Dependency graph analysis, AST transformations, and verification engineering do the heavy lifting. AI steps in to eliminate the manual effort that used to block teams for quarters.
Speed without understanding is just technical debt with a new name. PRExplainer uses Tree-sitter to build real context about the modified code, keeping it tight yet useful. That's why the explanations and reviews it generates are coherent, not hallucinated.
When you ship in days what used to take months, cloud costs sneak up on you. DevsBill tracks spending across AWS, GCP, and Azure with forecasting and optimization signals. Built for developers who move fast and pay their own bills.
The intelligence layer across all your projects. Indieprise surfaces cross-project signals: which codebases need migrations, how your code is evolving, where you're spending too much. One picture instead of five dashboards.
Also in the workshop
Multi-session command centers for developers juggling five Claude Code sessions at once. Autonomous application scaffolding. Local-to-cloud migration bridges. Most of what I build starts as a tool I needed yesterday.
Background
Fifteen years in rooms where the stakes are real. Investment banking tools, social products serving billions of users, systems handling private capital markets and portfolio valuations. Fintech, social platforms, healthcare.
That background shapes everything I build now. When I design an autonomous agent, I think about failure modes. When I ship a developer tool, I think about the on-call engineer at 2 AM. The instinct to build things that actually work in production comes from years of building things that had to.