His Rule From High-Frequency Trading for AI Agents
About this episode
I tried building Tabs, a $91M AI billing platform, from scratch with Claude and a cron job. CTO Deepak Bapat walked me through every layer I'd miss: context objects that push contract accuracy from 80% to the high 90s, the commercial graph no horizontal AI can replicate, and why older models sometimes outperform newer ones.
Key moments
- 0:00 What if AI could run your entire finance department?
- 1:08 How a high-frequency trader ended up building AI for finance
- 3:25 What does an AI billing platform actually do?
- 9:15 Are CFOs already using AI agents for reconciliations?
- 11:05 Should you build billing in-house or buy a platform?
- 13:36 What is a commercial graph and why can't horizontal AI build one?
- 17:54 Can AI agents actually replace human finance teams?
- 24:21 Will one person run finance for a $100M company?
- 27:07 How will AI agents from different platforms communicate?
- 28:54 Can you build a $91M AI product with Claude and a cron job?
- 35:23 Is AI accuracy a model problem or a context problem?
- 39:48 Why did this CTO delete the chat interface?
- 45:10 Are developers losing their coding skills because of AI?
- 46:42 Why invest in junior engineers when AI ships faster?
- 48:36 Should coding interviews be replaced with code review?
- 51:16 The book every interdisciplinary builder should read
- 53:08 outro