Learning by building
One post per week. What I am shipping, what I am figuring out, what is not working yet.
I built an AI sandbox so I could actually learn this stuff
Instead of reading about RAG and agents and embeddings in isolation, I built a place where they're all connected and I can break things on purpose.
Not everything should be automated. Here is how I decide.
The five questions I ask before building any automation. Frequency, judgment, cost of failure, maintenance, and adoption — and why skipping any of them leads to tools nobody uses.
GTM OS is getting an Agentic team
GTM OS started as a framework I built because I had no idea how to sell anything. Now it is getting seven AI advisors and four simulation agents — because a structured process is not enough if nobody is challenging your thinking.
I built a job discovery agent. Here is what actually happened.
The 504 error, the data deletion bug, why I chose Redis, and what building an AI agent in production actually looks like when things break.
One app and a plan I have no idea will work
I built CV Tailor for myself. Then it grew. Now I am using it to test a go-to-market framework I wrote from scratch — seven stages from zero to first ten paying customers. This is why I am building in public