Building with
AI, day by day.
Not a tutorial. Not a framework. A field manual from someone who actually ships — written session by session, with the people in the room.
"An LLM is a system that predicts the next token, trained on most of the internet. Every chatbot, every agent, every demo on your timeline — under the hood, this."
An LLM is a system that predicts the next token, trained on most of the internet.
Tokens, embeddings, transformers, self-attention. Math-free version of all the math that runs everything.
System prompts, few-shot, chain-of-thought. The craft of talking to LLMs.
Chunking, embedding, retrieval, re-ranking. Build a RAG pipeline from scratch.
From chatbot to agent. Function calling, orchestration, the agentic loop.
Conversation memory, user memory, world memory. Making AI products that actually know you.
How do you know it works? Evals, benchmarks, human preference. The boring part that matters most.
The $40K death spiral. Pricing AI products, token budgets, caching strategies.
Latency, streaming, fallbacks, monitoring. What running AI in production actually looks like.
Multimodal, reasoning models, agents at scale. Where all this is going.
Get each day when it drops.
No newsletter cadence. No sponsored content. Just a link when the next session is ready.