Azeem's agent habits and capabilities

Azeem describes R. Mini Arnold's skills: summaries, formatting preferences, and building defenses against prompt injection.

Play episode from 08:45

chevron_right

Transcript

chevron_right

Transcript

Episode notes

Welcome to Exponential View, the show where I explore how exponential technologies such as AI are reshaping our future. I've been studying AI and exponential technologies at the frontier for over ten years.

Each week, I share some of my analysis or speak with an expert guest to make light of a particular topic.

To keep up with the Exponential transition, subscribe to this channel or to my newsletter: https://www.exponentialview.co/
-----

In this episode, I sit down with my friend Rohit Krishnan - writer of the Substack newsletter Strange Loop Canon - for a hands-on conversation about what it actually looks like to build with AI agents today. Between us we're burning through tens of billions of tokens a month - I hit nearly 100 million in a single day this week - and we share what we're each running on our own machines.

We dig into the quirks and surprising power of tools like OpenClaw, Claude Code, and Cowork, debate why AI remains stubbornly bad at good writing, and zoom out to ask what a world of trillions of agents might actually look like — and what economic infrastructure it will need.

We covered:

(03:15) What's on your screen right now?

(04:30) OpenClaw

(06:27) Rohit’s agent, Morpheus

(11:06) Azeem's agent, R. Mini Arnold

(19:25) The analyst is now a machine

(22:36) 100 million tokens in a day: the new normal

(24:44) Building tools to improve AI writing: Horace and Broca

(32:19) Why writing is the hardest eval for LLMs

(39:18) Towards a trillion agents

(42:09) The agentic economy: coordination, identity, and exchange

(46:33) How to get started with OpenClaw