PM by instinct. Engineer by experiment. I work at the intersection of product and AI — deconstructing LLM systems, running experiments, and writing about what it means for how builders think and build.
Forward deployed // IndiaClaude Code, Codex, Cursor — strip them back and they're engineering products built on frontier APIs. This lab runs the experiments to understand what's inside, what's replicable, and what you can build on top of. Operator-first. Published when there's a real finding.
Built from first principles for research that actually holds up. Multi-source web search, structured orchestration, context-engineered for signal over noise. AI research should be verifiable, not just plausible.
An expert version of yourself — contextually aware and up to date — without giving up your data to build it. Your knowledge, your context, your control. Bring your own key. No black box.
Attention is the input. Train deliberately or the algorithm trains you.
The model changes every six months. Build the process first — then let tools serve it.
Systems need receipts. A plan without a feedback loop is just optimism.
Open by Default
Sovereignty starts with open systems. Design for interoperability, not lock-in. Build for the user, not the platform.
Conscious Design
Every decision is intentional. From architecture to copy, we ask: Does this serve the user, or the system?
Sovereignty as a Right
Your data, your knowledge, your tools. No black boxes. No hidden costs. Only transparency and control.