← newsletter

Originally published on 12 Grams of Carbon

Agentics: understanding the agent stack

Understanding the agent stack from the user’s keyboard all the way to the datacenter and back

Gaurav · August 12, 2026

Editors Note: We’re one week out from the August Agentics Flagship Event! If you’re in NYC next Wednesday, come out to midtown for a stacked night with free dinner. We have seven talks from people who build agents for a living sharing what’s working today. Modal, Warp, Goose, Librechat, Opus, Subconscious, and Nori. Register if you haven’t already.


This is a guest post from Gaurav, our GTM Intern at Nori Agentic. Gaurav’s a rising junior at NYU studying Business and Philosophy, but I know him because he went to my high school a decade after me, which makes me feel really really old. Gaurav initially reached out to me over Substack DM about one of my AI blog posts, which apparently was just an “in” to get me to speak to his club at NYU after I ghosted him on email. In other words, a great GTM leader in the making.


I’m not technical, at all. My extent to being technical is taking Data Structures last semester (but the teacher only tested our code and never read it/compared it to Claude’s outputs … you know how that went). So you can imagine when I joined Nori and had to figure out how to market “background agents”, with no clue how regular agents even worked, I had a steep hill to climb.

I recently had a teaching session with our CTO Cliff, and he walked me through his new blog post of “what’s in the agent stack.” Being non-technical, I thought that the diagram he drew up on the blog post made perfect sense, and it really helped me visualize what’s actually in this black box of an “ai agent”. I thought his blog could help a lot of other people, but Cliff hasn’t actually posted it anywhere (Cliff I know you read these, you should post it)

Anyway, I thought it would be cool to create an interactive web walk through of the blog post, sorta in the style of old-school web projects. Yes, still non-technical, but turns out you can get pretty far by just asking the nori bots to keep going and going back and forth on random UI polish. Literally never been easier to build cool things!

Content spans from easy to digest for beginners to also informational for more experienced people. Check it out and feel free to comment your thoughts or reach out to me directly - gaurav@noriagentic.com (would love to hear from you!)

Full site at https://the-agent-stack.com/.

Landing screen of the-agent-stack.com
welcome to my house ill be your guide
Tool calling inside the agent harness
deep dive on tool calling inside the agent harness (I know what these words mean now!)
Following a single request up and down the agent stack
built the site you can follow a single request all the way up and down the stack. little Terraria design here that I thought would be cool
The context layer of the agent stack
context layer part of the stack
The hyperscaler layer of the agent stack
finally learned what a hyperscaler is
The full agent stack
full stack!

I’m planning to write a more full blog post that is just about the experience of actually building this site as a totally non-technical person. I guess the tldr would be, it’s really crazy how easy it is to build things now.