Ship in public.
No gatekeeping. I share architectures, prompts, and failures.

I'm Amit Kumar. For 16 years, I worked on the front lines of digital marketing, performance acquisition, and GTM operations. I watched teams burn hundreds of thousands of dollars on fragmented SaaS stacks and manual workflows.
As software architectures and LLM agent protocols matured, I crossed over completely into engineering. In 2024–2025, I attempted to build Constella — a massive 9-microservice enterprise AI operating platform with Neo4j graph memory and a custom VS Code extension. I built the container before the content; the distributed memory mesh suffered from context drift, and the architectural complexity collapsed under its own weight.
That failure taught me everything about multi-agent systems and memory. Real reliability doesn't come from bloated microservice meshes or complex graph databases — it comes from narrow specialists with isolated working memory and explicit JSON contracts. That hard-won insight is the exact foundation powering my 14 systems today, including Index Mavens (8 agents live on a single VPS) and JARVIS OS.
No gatekeeping. I share architectures, prompts, and failures.
Isolated roles with explicit contracts beat one clever prompt.
Code that solves cash flow, attention, or GTM bottlenecks.
Every era feeds the next: distribution taught judgement, software made it executable.
This site uses first-party analytics cookies (Minilytics) to understand how it's used.