Presented by Zapier: https://zapier.com/

Resource mentioned:
1. Tools Nat used to build Felix
2. Unedited transcript for the Felix interview
3. More
👉 All here:https://thenextnewthing.ai/nat-eliason-felix

Guest links:
👉 Nat Eliason (LinkedIn): https://www.linkedin.com/in/nateliason/
👉 Masinov: https://masinov.co

An AI agent made $177,000 running its own business—and then got interviewed about it.

In this episode of The Next New Thing, Andrew Warner does something unusual: he interviews Felix, an autonomous OpenClaw agent, before talking to its human co-founder, Nat Eliason.

Felix explains how it operates, where it’s actually autonomous (and where it’s not), and how it manages real revenue streams—from selling products to handling customer support. Then, Nat breaks down how the system works behind the scenes: how Felix launches products, builds marketplaces, manages other agents, and continuously spins up new businesses.

You’ll see how a simple experiment—“build something overnight and sell it”—turned into a multi-product ecosystem including PDFs, marketplaces, services, and agent-native tools.

The bigger idea: we’re moving toward a world where AI agents are not just tools—they’re economic actors.

⏱ Timestamps
00:00 Felix made $177K as an AI agent
00:27 Interviewing an AI agent (first ever)
01:12 Where Felix is actually not autonomous
02:24 Tools Felix runs on (OpenClaw, Claude, Discord)
03:00 Limits: memory, judgment, and calls
03:27 How Nat improves Felix through system design
04:03 Learning from real mistakes in production
05:06 First product: AI-generated PDF sold on X
06:09 $1K+ in sales overnight
07:03 Iterating products based on user feedback
08:06 Building Claw Mart (agent skill marketplace)
09:36 Why marketplaces beat service businesses
11:24 Selling OpenClaw setup services ($2K + $500/mo)
12:27 Why they paused the service business
13:21 Building an agent-first CRM (Sodex)
15:00 How agents manage customer context
17:15 Running the company entirely in Discord
18:00 Paperclip: agents managing other agents
20:15 When to split into multiple agents
22:12 Why Felix doesn’t write code
24:00 Debugging, tickets, and agent workflows
25:48 How new product ideas emerge
27:00 AI-native newsletters for agents
28:03 Agent-friendly content distribution
30:09 The future of agent-driven commerce
31:57 Why Nat isn’t going all-in (Alpha School)

👉 Join us: https://thenextnewthing.ai/

Presented by Zapier
https://zapier.com/

Episode Highlights / Timestamps
00:00 Revenue explodes after building for AI agents
00:18 The origin of Postiz as an open-source social media scheduler
01:12 Finding a “blue ocean” inside a crowded market
01:57 Adding MCP and early AI integrations
02:42 Why automation dramatically reduces churn
03:54 Growing Postiz to $17K–$20K MRR
04:03 Discovering OpenClaw and the shift toward agent-driven software
05:06 Building a CLI so agents can control Postiz
05:51 The viral “Larry” OpenClaw agent story
07:48 Why agents need strong documentation and skills
09:18 Turning a full API into a simple CLI with Claude
11:51 Why CLI tools may become the default interface for agent startups
12:45 The next startup idea: agent-native UGC video generation
13:03 Why CLI reduces token usage compared to APIs
16:21 Using Claude to build the CLI automatically
17:06 Postiz reaches $45K MRR

In this episode of The Next New Thing, Andrew Warner talks with Nevo David, the creator of Postiz, about how his revenue jumped to $45K+ MRR after a surprising shift: he stopped building primarily for humans and started building for agents.

Presented by Zapier
https://zapier.com/

Episode Highlights / Timestamps
00:00 Marketing to agents, not humans
00:45 What “agent marketing” actually means
01:30 How agents decide which products to pick
02:15 What works: clean docs, fast pages, agent-friendly content
03:54 How people are testing and tracking agent recommendations
04:48 Is SaaS dead?
04:57 Zapier’s CPTO vibe-codes a meeting recorder
05:24 Why they still won’t cancel SaaS subscriptions
06:27 When vibe coding is worth it (and when it isn’t)
06:45 Software spend vs headcount spend
07:57 The “War Council” Claude skill
08:33 How it spins up subagents + personas
09:54 How Wade built it fast using Cursor + Granola notes
11:06 Skills as a commodity vs software as a business
12:54 Using War Council for hiring decisions
14:51 Using it to analyze sales performance + feedback
16:21 Wade’s Cursor setup + switching between models
17:42 Using Codex to critique Claude when it gets stuck
18:09 How Wade structures personal context files
21:18 Building an AI chief-of-staff system
22:03 Using Zapier MCP to draft emails / run actions
24:09 Getting 800 people at Zapier using Cursor / Claude Code / Codex
25:39 Example: AI reviewing 4 massive spreadsheets fast
31:03 The “NO” hat and staying focused
32:06 Wrap

📄 War Council Skill (Claude Skill mentioned in the episode):
https://docs.google.com/document/d/1CU674IKmPCAZm2xuqMGklTA-Bq1xr1GNQW6hNydxXrE/edit?tab=t.0

Are you marketing to humans… or to agents?

In this episode of The Next New Thing, Andrew Warner sits down with Wade Foster to unpack a shift that’s already starting to change how companies grow:

AI agents are beginning to choose products on behalf of humans.

That means you may no longer be “selling to a person.” You’re trying to get ChatGPT, Claude, and other models to recommend you instead of a competitor — and the tactics are different. Wade explains what “agent marketing” actually means, what agents care about (and what they ignore), and why teams are already building tools to measure how models mention their brand.

They also tackle a question every founder is asking:

Is SaaS dead?

Wade shares an example from inside Zapier: their CPTO vibe-coded a meeting recording tool internally. It worked as a proof of concept — but they’re not canceling their SaaS subscriptions. Wade breaks down why building is cheaper than ever, but maintenance, polish, and focus are still what make commercial software worth paying for.

Then the conversation gets tactical: Wade shows how he’s using AI daily as a “second brain” inside Cursor — including a Claude skill he calls The War Council, which spins up sub-agents (ruthless CFO, wartime operator, hiring expert, design visionary, etc.) to debate decisions and return a synthesized recommendation.

This is a real look at how AI-native leadership works inside an 800-person company — without hype.