Greg Isenberg

Making $$ with AI Agents

3 min read30 Apr 2026Worth watching

TL;DR

Greg Isenberg interviews Howie Liu, co-founder and CEO of Airtable, about the massive opportunity in AI agents and demos Hyperagent.com — a Mac-like, UX-focused AI agent platform that lets users build, deploy, and manage fleets of digital employees for a fraction of human labor cost.

Key points

Howie Liu argues the true TAM for AI agents is not $1 trillion but tens of trillions — essentially all white-collar labor GDP — because frontier models are now smart enough to autonomously execute complex multi-step tasks across every industry

Hyperagent.com is positioned as the Mac to OpenClaude's Linux: cloud-native, UX-focused, with built-in tools like Google Maps, image generation, Slack deployment, and full coding capability powered by frontier models like Opus 4.5

The platform features Skills (reusable, self-improving agent playbooks), Rubrics (LLM-as-judge eval scoring for quality oversight), and a command center fleet view — enabling one person to manage and oversee dozens of specialized agents

Howie demonstrated building a full business case AND a working v1 app for a hyperlocal real estate market report idea entirely through Hyperagent, including Reddit validation, competitive analysis, and a clean UI — treating the agent as a founder, not just a developer

First 1,000 people to sign up get $1,000 in free Hyperagent credits (Howie committed $1 million total in tokens to the Startup Ideas podcast community)

Actionable insights

→

Commit to using a frontier agent product like Hyperagent every single day for 30-60-90 days — sporadic use is why most people underestimate AI agents and fail to see compounding returns

→

Stop anchoring AI cost to $10/month SaaS subscriptions — compare token cost to human equivalent time cost instead (e.g., $150 for a board memo that would take days and come out worse is a bargain)

→

Build Skills in Hyperagent by having it research your actual style and voice, then pin Rubrics to auto-score every output with an LLM judge — this creates a scalable self-improving content or research system without manual review of every output

→

Use Hyperagent's Slack deployment to make agents always-on virtual coworkers that listen to channels and chime in with relevant expertise — one click from any agent you build

→

Do not one-shot a prompt and give up — agents require coaching, iterative feedback, and skill refinement the same way a new human employee would need onboarding and correction

Notable quotes

“The TAM for that is like not even a trillion. It is like probably like the whole GDP of like all white collar labor which is like obviously many tens of trillions.”

“A lot of it was researched and crafted by Hyper Agent and I got feedback that that was the best memo from some of our best investors that I had ever written. And I am like yeah because an agent did it.”

“It is kind of like management 101 but applied to agents now where it is like as you scale up if you are the CEO of a business you just literally do not have time to go and look at every single thing that every single person in the company has done.”

Worth watching?

✅

Worth watching the full video?

Watch if you want to see Hyperagent demoed live with real business use cases — the key concepts, platform features, and strategic framing are all captured here, but the screen-share walkthrough adds useful visual context.

Topics

AI & Tech Airtable

Explore more summaries on these topics →

Saved you some time? The creator still deserves a like.

Watch on YouTube →

More like this