Why I Wrote Intelligence at Scale

The questions I encounter while building AI products rarely stay inside one component.

A conversation about a model becomes a conversation about retrieval. A tool integration raises questions about permissions. A working demo raises a harder question: what evidence would convince the team that it is ready for real users?

That is the territory of my book, Intelligence at Scale: LLMs, agents and the systems around them.

I am Shashank Agarwal, founder and CEO of Noveum AI and creator of API.market. Writing has been part of how I learn since I started publishing technical tutorials while in school.

Connect the parts of the system

It is possible to understand an individual API without understanding the agent that calls it. It is also possible to build an agent without knowing which failures its evaluation catches.

A useful learning path should help the reader connect those parts. What is the agent trying to do? What information can it use? Which actions can it take? How do you inspect a run and decide whether the outcome met the task?

These questions turn an interesting demonstration into an engineering discussion.

Use diagrams to expose a decision

A diagram helps when it makes a relationship easier to inspect.

For an agent workflow, that might mean showing where a tool is called, where a result enters the context, or where a human reviews an action. The diagram should make an assumption visible rather than decorate the surrounding text.

The book's diagram collection is available with its stated license. Readers and speakers should check that license before reusing a diagram.

Give the reader something to check

When I read a technical explanation, I want to know what it enables me to do differently.

After a discussion of tool use, I should be able to inspect a small workflow. After a discussion of evaluation, I should be able to describe a failure the test would catch. After a discussion of deployment, I should know which uncertainty remains.

That is the standard I want technical writing to serve.

The book and my companies explore connected parts of that work: API.market helps developers and agents access tools, while Noveum focuses on production-agent evaluation and investigation.

I share my writing and product work on LinkedIn.

Author's introduction, drafted with AI assistance. See the book site for the current edition and availability.