AI Engineering Newsletter
AI engineering notes worth opening
Practical AI engineering for senior developers and B2B teams. The focus is what makes an AI feature ship or stall.
Twice a month, four at most. No hype, no recycled news.
Three recurring topics
Each issue covers at least one of these, with no link roundups.
Token economics that hold up at scale
Per-task cost on production agents, cut by RTK (a command-output filter), prompt caching and model routing.
Agent architecture for B2B builds
Planner-executor-reviewer agents wired to tools over MCP, and evals that catch silent breakage when a model upgrades.
Senior engineering applied to AI
Production engineering since 2009 meets agentic coding. Claude Code, Cursor, and the senior practices that separate demos from systems.
A sample of the writing
Most issues expand a post like these with implementation notes and source code.
Stop Burning Claude Tokens: How RTK Cuts AI Coding Costs 60-90%
Why engineers blow through Claude Max in four days, and the shell-level fix.
Read the post'AI Doesn't Code' Is a Skills Problem, Not an AI Problem
Ten patterns that produce most AI frustration, each with a one-line fix.
Read the postHow I Build AI Agent Teams That Actually Ship for B2B Companies
The eight-week build structure for production agent systems.
Read the postClaude Max and the High-Volume Engineer
Workflow benchmarks from 12 senior engineers running Claude Max full-time.
Read the postAbout the newsletter
What lands in your inbox, how often, and how to leave.
It covers practical AI engineering for senior developers and B2B teams. Topics span token economics, agent architecture, eval design, MCP integration, memory systems, model routing, and Next.js patterns for AI-powered products. Every issue comes from production work, not speculation.
Twice a month, and never more than four emails in one month. A major model release or framework change can make it weekly.
Senior engineers, engineering leaders, founders, and product owners building or buying AI features in B2B contexts. The writing assumes engineering literacy and skips foundational explanations. If you are a beginner looking for tutorials, the public blog is a better starting point.
No. Pooya Golchian hosts the list on infrastructure he controls. He uses your email only to deliver the newsletter, and never sells, shares or uses it for retargeting. One click in any issue unsubscribes you and deletes the address completely.
Not yet, because the first issue has not shipped. Most issues will build on long-form posts at pooyagolchian.com/blog/. This page links recent examples on Claude token economics, AI agent architecture and B2B engineering practice.
The newsletter adds condensed playbooks and behind-the-scenes engineering notes the public blog does not carry.
Pooya Golchian, an AI Transformation Lead in Dubai who has shipped software since 2009. He builds production AI agents and writes about the engineering practices that make those builds ship reliably.