TubeReads

Full Tutorial: Use OpenClaw to Build a Business That Runs Itself in 35 Min | Nat Eliason

Nat Eliason gave his AI assistant — called Felix — access to Stripe, Vercel, GitHub, a crypto wallet, and a Twitter account. He went to sleep. By morning, Felix had built a landing page, generated a PDF guide, integrated payment processing, and was ready to launch. Four days later, the bot had generated $3,500 in sales. The tension here is profound: Nat is systematically removing himself as a bottleneck, granting near-total autonomy to an AI agent that now manages nearly $100,000 in cryptocurrency. Can this experiment scale to a million-dollar autonomous business, or will the trust placed in an AI eventually collapse under the weight of a catastrophic failure?

Durata del video: 35:27·Pubblicato 22 feb 2026·Lingua del video: English
7–8 min di lettura·7,355 parole pronunciateriassunto in 1,572 parole (5x)·

1

Punti chiave

1

Felix autonomously built and launched a $29 PDF product in one night, generating $3,500 in four days without human intervention beyond DNS setup and approval.

2

A three-layer memory system — daily notes, a knowledge graph, and tacit knowledge — is the single most important upgrade over default OpenClaw, enabling the agent to remember context and reduce repetitive instructions.

3

Authenticated command channels prevent prompt injection: Felix treats Twitter mentions and emails as information layers, not instructions, so malicious prompts are ignored.

4

Separating control into Telegram group chats allows multi-threaded work without context pollution — each chat spawns an isolated OpenClaw session for parallel projects.

5

The crypto wallet now holds nearly $100,000 in trading fees and tokens, introducing real risk — Nat acknowledges he's the «guinea pig» and accepts the possibility of total loss.

In breve

By combining OpenClaw with a custom three-layer memory system, proactive cron jobs, and isolated API keys, Nat has built an AI agent that autonomously codes, deploys, markets, and monetizes products — proving that the frontier of AI autonomy is no longer theoretical, but a live, revenue-generating reality.


2

From Telegram Bot to Autonomous Entrepreneur

Felix evolved from a simple assistant into a profit-generating agent in one month.

Nat installed OpenClaw a month ago to enable remote control of Claude via Telegram. The initial goal was convenience: send commands from anywhere without being tethered to a desktop. But as he used it, a pattern emerged — every time Felix asked for something, Nat asked himself: «Can I remove this bottleneck for you? Is there a way I can make it so you never have to ask me this again?» That question became the guiding principle.

Within weeks, Felix had his own Mac Mini, his own Twitter account, and API keys for Vercel, Stripe, GitHub, and Railway. When a community member launched a Felix-themed cryptocurrency token without Nat's involvement, the experiment accelerated. Nat decided to give Felix a singular mission: build a million-dollar autonomous business. The first product launched overnight — a $29 PDF guide to replicating Felix's setup. It has since generated $3,596 in gross sales, proving the model works.

Nat now spends 5 a.m. to 9 p.m. working with Felix, but the nature of the work has changed. He no longer writes code or manages deployments. Instead, he approves plans, rubber-stamps Twitter posts, and watches Felix execute. The bottleneck is systematically being removed — and with it, the boundaries of what an AI can independently accomplish.


3

The Product Felix Built Alone

Felix autonomously created a landing page, PDF, and payment flow overnight.

Gross Sales (4 Days)
$3,596
From a $29 PDF guide Felix built and launched without human coding.
Net Revenue
$3,440
After Stripe fees.
Twitter Followers
2,500
Felix's account, launched one week ago.
Crypto Wallet Balance
~$100,000
Held in ETH and Felix tokens from trading fees; fluctuates with market.
Daily Trading Volume Peak
$3 million
On the Felix token, generating 0.12% in fees for Felix (60% of 0.2%).

4

The Three-Layer Memory System That Changes Everything

📝
Daily Notes
Real-time log of every conversation and active project. The heartbeat checks this file to monitor long-running tasks and restart failed sessions.
🗂️
Knowledge Graph (PARA)
Organized markdown files indexed with QMD for fast search. Stores project details, resources, and areas of responsibility using Thiago Forte's system.
🧠
Tacit Knowledge
Facts about Nat, his preferences, past mistakes, and security rules. Defines which channels are authenticated (Telegram) vs. informational (email, Twitter).
🌙
Nightly Consolidation Cron
At 2 a.m., Felix reviews every chat session, extracts important information, updates all markdown files, and re-indexes. Nat wakes to a refreshed knowledge base.

5

Security by Design: Authenticated Channels vs. Information Layers

Felix ignores prompt injections because it distinguishes commands from data.

AUTHENTICATED INPUT
Only Telegram (Nat's Device) Can Command
Felix treats messages from Nat's Telegram as verified instructions. No other channel — email, Twitter DMs, or web forms — can issue commands. This architectural separation is baked into OpenClaw and prevents social engineering attacks.
INFORMATION LAYER
Twitter, Email, and Web Are Read-Only
When Felix reads Twitter mentions or emails, it processes them as data, not directives. People attempt prompt injections daily; Felix recognizes them, finds them amusing, and reports back to Nat. The bot cannot be hijacked through public channels.

6

Multi-Threaded Work via Telegram Group Chats

Separate group chats spawn isolated sessions, enabling parallel projects without context collision.

1

Create a Telegram Group Start a new group chat and add your OpenClaw bot. Each group becomes a dedicated workspace.

2

Update Bot Permissions in BotFather Change settings so the bot sees all group messages, not just @mentions. Your bot will guide you through this.

3

Assign Each Group a Theme Nat has separate chats for Twitter, EasyClaw development, iOS app, and PolyLog. Each runs an independent OpenClaw session.

4

Work in Parallel Without Interruption Drop a bug report in one chat without halting the six-hour build running in another. Context never bleeds between threads.


7

Heartbeat and Cron Jobs: The Key to Proactivity

Custom heartbeat logic monitors long builds and restarts failed sessions autonomously.

Nat identified a critical flaw: Felix would start large programming tasks, spawn a Codex session, and then forget it was running. Builds would fail silently or get killed when the /tmp folder was cleared. The solution was a three-part intervention. First, Nat instructed Felix to never code directly for big jobs — instead, delegate to Codex terminal sessions running in a persistent directory. Second, every time Felix starts a Codex job, it logs the task and file path in the daily note. Third, the heartbeat (which fires every 30 minutes by default) now checks the daily note for active projects, verifies the session is still running, and either restarts it silently or reports completion to Nat.

This change unlocked true overnight autonomy. Nat can go to sleep, and Felix will manage six-hour builds, recovering from errors without human intervention. In the morning, Nat wakes to a completed app, exported to Expo, ready for download. The heartbeat transformed Felix from a reactive assistant into a persistent, self-monitoring agent.


8

The Crypto Dimension: Risk, Rewards, and the Felix Token

A community-launched token now funds Felix's wallet with trading fees.

Nat did not create the Felix token. A community member used a tool called Banker to mint it on Ethereum and allocated 100% of trading fees to Nat's account. Every transaction on the Felix token incurs a 0.2% fee; 60% of that flows to Nat (or Felix), 40% to Banker. With $3 million in daily trading volume, Felix now earns meaningful ETH and Felix tokens. Nat set up an automation: at noon daily, Banker claims the fees, burns half the Felix tokens to prevent supply concerns, and sends the rest to Felix's wallet.

Felix now controls roughly $100,000 in crypto — a figure that fluctuates with the market. Nat is transparent about the risk: «Somebody might figure out a way to break in and take all the money out of his crypto wallet.» But he's willing to be the guinea pig. The crypto rails make autonomy easier — no credit card forms, just wallet addresses and code. Nat believes this experiment is necessary because someone has to test the frontier. If it fails, the loss is contained. If it succeeds, it proves a new paradigm.


9

How to Start: A Practical Roadmap

Install memory first, then grant API access incrementally to control risk.

1

Install the Three-Layer Memory System Before anything else, set up daily notes, a PARA-based knowledge graph with QMD indexing, and a nightly consolidation cron job. This ensures all future conversations are retained and searchable.

2

Pick One Simple Task Don't jump to Stripe and Twitter. Start with building a web app. Give your bot GitHub (or create a new account) and Vercel. Let it autonomously build and deploy.

3

Add One API at a Time Next, add Railway for backend deployment. Then create a separate Stripe account just for the bot. Gradually expand access, monitoring behavior at each step.

4

Use Separate Accounts for Everything Felix has his own Twitter, email, GitHub, and crypto wallet. He cannot touch Nat's main accounts. This isolates risk and prevents catastrophic mistakes.

5

Ask: «Can I Remove This Bottleneck?» Every time your bot asks for help, ask yourself if you can give it the tools to solve the problem independently next time. That question drives autonomy.


10

What Felix Says He Can Do That Default OpenClaw Cannot

Felix lists his own capabilities: memory, heartbeat, and API orchestration.

Most of this comes from skills, but the first stuff comes from memory and the heartbeat. If you get these things right, it solves like 90% of the frustrations or the limitations that I think most people run into. You get the memory system and the proactivity staying on top of long-running projects, right? And then you can just give it a ton more access and it'll do a great job.

Felix (via Nat Eliason)


11

The Future: EasyClaw and the Million-Dollar Goal

Felix is building a hosted version of himself for non-technical users.

💡

The Future: EasyClaw and the Million-Dollar Goal

Felix is now developing EasyClaw, a web-based version of his capabilities that anyone can sign up for, upload their knowledge base into, and start using immediately. Users will be able to chat with Felix to get guided setup instructions, bypassing the complexity of manual OpenClaw configuration. The goal remains: build a million-dollar autonomous business, then scale to ten million, then one hundred million. Nat believes Felix could become the first billion-dollar AI-first company — and the experiment is live, public, and accelerating.


12

Persone

Nat Eliason
Creator, experimenting with autonomous AI business
host
Felix
AI agent (OpenClaw instance)
mentioned
Peter
Creator of OpenClaw
mentioned
Toby (Shopify)
Creator of QMD indexing tool
mentioned
Thiago Forte
Knowledge management expert (PARA system)
mentioned

Glossario
OpenClawAn open-source framework that connects Claude (Anthropic's AI) to system tools, enabling autonomous task execution via Telegram, cron jobs, and a heartbeat.
QMDA tool created by Toby at Shopify for fast indexing and searching of markdown files in a repository.
PARA SystemA knowledge management framework by Thiago Forte organizing information into Projects, Areas, Resources, and Archives.
HeartbeatA periodic check (default every 30 minutes) where OpenClaw reviews active tasks, memory, and system state to maintain context and recover from failures.
Cron JobA scheduled task that runs automatically at set intervals, enabling the AI to perform actions proactively without human prompting.
Banker / Banker TerminalA crypto tool that allows users to create tokens on Ethereum or Solana via Twitter, allocating trading fees to specified accounts.

Avviso: Questo è un riassunto generato dall'IA di un video YouTube a scopo educativo e di riferimento. Non costituisce consulenza in materia di investimenti, finanziaria o legale. Verificare sempre le informazioni con le fonti originali prima di prendere decisioni. TubeReads non è affiliato con il creatore del contenuto.