Day 1 of 28 · OpenClaw Challenge
What Are AI Agents?
⏱ 5 min
📊 Beginner
You've probably used ChatGPT. You type something, it responds. You type again, it responds. That's a chatbot — it talks, but it can't do anything in the real world.
Today you'll learn why AI agents are fundamentally different — and why they matter.
Knowledge Check
When you ask ChatGPT to "schedule a meeting with Sarah for Friday" — what happens?
A
It connects to your Google Calendar automatically
B
It creates the calendar event and sends Sarah an invite
C
It writes you a message you could copy-paste into your calendar app
Exactly. ChatGPT can only give you text. It can't touch your calendar, your email, or any of your tools. That's the core limitation of a chatbot.
Let's see what this looks like in practice. Here's what happens when you ask a regular chatbot to help with your inbox:
I have 200 unread emails. Can you help me clear my inbox?
I'd be happy to help you organize your inbox! Here are some strategies:
1. Start by filtering by sender
2. Unsubscribe from newsletters you don't read
3. Use the "star" feature for important emails
4. Set up filters for recurring emails
Would you like me to draft template responses for common email types?
No, I mean can you actually go through my emails and sort them?
I'm sorry, I don't have access to your email inbox. I can only provide advice and draft text for you to use. You'll need to manually sort through your emails yourself.
Would you like me to help you create a sorting system?
↻ Replay conversation
Knowledge Check
What was the chatbot's problem?
A
It wasn't smart enough to understand the request
B
It needed a better prompt
C
It has no connection to your email — it can only give advice
No amount of clever prompting fixes this. The chatbot simply can't access your tools. It lives in a text box, disconnected from everything else.
Now watch what happens when you send that same request to an AI agent — one that's actually connected to your email:
I have 200 unread emails. Can you help me clear my inbox?
On it. Let me scan through your inbox...
Done. Here's what I found across your 200 unread:
🗑 47 promotional emails → moved to trash
📂 38 newsletters → moved to 'Read Later'
⏳ 12 automated notifications → archived
✉️ 6 emails need your reply
⭐ 3 urgent from your team
I've left the important ones in your inbox. Want me to draft replies for the 6 that need responses?
Yes please, draft replies for all 6
Drafts ready ✓
I've saved them in your Drafts folder. Here's a summary:
1. Sarah (meeting request) → confirmed Friday 2pm
2. Tom (project update) → asked for deadline clarification
3. Finance (expense report) → approved with note
4. Client (proposal feedback) → thanked + scheduled follow-up
5. HR (policy update) → acknowledged
6. Dev team (bug report) → assigned to Jake
Review them when you're ready, or say 'send all' and I'll fire them off.
↻ Replay conversation
Knowledge Check
What did the AI agent do that the chatbot couldn't?
A
It used a more advanced AI model
B
It actually accessed the inbox, sorted emails, and drafted real replies
C
It gave better advice about email management
This is the fundamental shift. The agent didn't just talk about email — it did the work. Same AI brain, but with hands to act.
A chatbot can only generate text. An agent connects to your tools, remembers context, and takes real action.
The key differences
Think of it this way:
A chatbot is like texting a really smart friend. Great advice, but they can't touch your stuff.
An agent is like hiring an assistant. It connects to your real tools, remembers your preferences, works in the background, and takes action on your behalf.
Knowledge Check
Which of these could an AI agent do but a chatbot could NOT?
A
Write a poem about your weekend
B
Explain quantum physics in simple terms
C
Monitor your Slack and summarize messages you missed while in a meeting
D
Translate a paragraph from English to Spanish
Monitoring Slack requires a connection to Slack and the ability to act autonomously. The other three are just text generation — any chatbot can do those.
What real people are doing with OpenClaw right now
These aren't hypothetical. These are actual OpenClaw users:
🚨 Sentry alert: TypeError in /api/checkout line 142. Looks like a null reference when user has no saved payment method.
I've investigated. The bug is in handlePayment() — it assumes user.paymentMethods is always an array, but it's null for new users.
Fix: added a null check with fallback to empty array.
Tests passing ✓
PR opened: github.com/yourapp/pulls/847
Nice catch. Looks good, merge it.
Merged and deployed to staging. I'll keep monitoring for the next hour.
↻ Replay conversation
Knowledge Check
In that example, when did the developer first ask the agent to look at the bug?
A
They didn't — the agent detected it on its own and started working
B
They set up the task the night before
C
They asked after getting a notification
That's the "proactive" part of an AI agent. It doesn't wait for you to ask. It monitors, detects issues, and acts — then reports back to you.
Why this matters right now
Every major tech company — OpenAI, Google, Anthropic, Apple — is building toward AI agents. OpenAI just hired the creator of OpenClaw.
But here's the thing: OpenClaw is open-source. You can run it today, for free, on your own hardware. You don't need to wait for a company to give you access or pay a monthly fee for the AI to act on your behalf.
Andrej Karpathy, former AI Director at Tesla, called it the most incredible thing he'd seen recently. Over 190,000 people have starred it on GitHub. And the community is building new capabilities every day.
Good morning! What's on my plate today?
Morning! Here's your day:
📅 3 meetings (first one at 10am with Design team)
✉️ 2 emails need your reply — I've drafted responses
📊 The Q1 report you asked me to pull is ready in your Drive
⚠️ Heads up: Sarah rescheduled your 1:1 to Thursday
Anything you want me to handle before your first meeting?
Send those email drafts and clear my afternoon — I need focus time
Emails sent ✓
I've moved your 2pm and 3:30pm meetings to tomorrow and blocked 1–5pm as 'Focus Time' on your calendar. Both attendees have been notified.
You're free after lunch. I'll hold your messages unless something urgent comes in.
↻ Replay conversation
Final Check
After this challenge, you'll be able to set up an AI agent that can do which of these?
A
Connect to your real tools and take actions on your behalf — like managing email, calendar, and workflows
B
Replace all your software with one AI app
C
Only answer questions in a chat window
That's the goal. Over the next 27 days, you'll go from zero to having your own AI agent handling real work for you.
🎉
Day 1 Complete
"A chatbot talks. An agent does. In 27 days, you'll have your own."
Tomorrow — Day 2
Meet OpenClaw
We'll explore exactly what OpenClaw is, how it's built, and what makes it different from every other AI tool.