Welcome back, Creators! 👋

We've got +332 new builders joining our community this week.

Fresh from a week's digital detox, we're back with our learnings.

Let's jump in...

🔎 Preview: In today's issue

• 🤔 What is "Chain-of-Thought reasoning"?
• 📺 AI Agents 101 with Cien Solon
• 🚀 Lindy.io has cooked up something impressive

🤓 Explain Like I'm 5
​

Think of how you solve a complex problem in real life: First you break it into steps, work through each piece, then combine insights for a final answer.

Up until recently you needed to ask your AI to approach problems like this for the best results.

Thankfully, now AI models like GPT-5 and Claude 4 now do this automatically. They can perform internal "chain-of-thought" reasoning, working through problems step-by-step before responding.

🧠 The mental model

Before, AI was like a Magic 8-Ball - shake it, get an answer, with no visibility into how it got there. Now it's like watching a moderately intelligent intern (ouch!) think aloud, showing each logical move.

GPT-5's router decides when to think for 10+ seconds on hard problems versus responding instantly. Claude and others can extend reasoning also. Both deliver more reliable answers because they can "think before they speak."

🤓 A Brief History
​​
Prompt engineers began using chain of thought prompting as a way to unlock deeper reasoning in LLMs. By explicitly asking the model to break down its thinking, they found it could tackle harder problems and provide more reliable, audit-friendly answers.

⛓️ Chain of Thought Prompting

While this has nearly been made obsolete by specific reasoning models; you can still steer the LLM to behave how you want it to.

Instead of "What's the best pricing strategy?"—prompt step-by-step: "First, analyse competitors. Next, examine costs. Then factor user feedback..."

☕️ Knowledge Shots

1️⃣ Trigger even deeper thinking: Say "think step-by-step" or "show your reasoning", "think BLOODY hard".

2️⃣ Audit the steps: If the work merits it, review those steps to spot flawed assumptions.

3️⃣ Track the costs: Claude Opus 4.1 can be expensive; so think about how valuable the answer is to you.

🧠 So what next...

Test it yourself:

• Ask: "Should I use Bubble.io, Lovable or Flutterflow for my startup?"
• Then ask (or select 'thinking mode' on ChatGPT-5: "Should I use Bubble.io, Lovable or Flutterflow? Walk me through the decision factors step by step."

or see if ChatGPT realises you want a more reasoned response by itself.

📺 VIDEO: AI Agents 101

​

AI agents are often confused with automation, generative AI, or even AGI - but what’s the real difference?

In this session from Create With 2025, Cien from Launch Lemonade breaks down the key concepts, from the rigid workflows of traditional automation to the adaptive, tool-using nature of AI agents.

Drawing on real-world examples and lessons from building 10,000+ assistants, she offers a clear framework for understanding where we are today and where agentic systems are heading.

A practical, no-nonsense introduction for anyone trying to make sense of the hype and start building with AI agents.

▶️ Watch Video​

🔊 On your radar

Keep your finger on the pulse of what we think is hot in the space right now 🔥

⭐️ Airtable’s AI Pivot: Howie Liu’s Radical Rebuild ​

👉 Why it matters: Airtable didn’t just add AI—they tore up the org chart and rebuilt from scratch. The CEO codes daily, teams ship faster, and the product’s now a living, learning system. They're operating with a 'fast' and 'slow' team structure which means, where it makes sense, teams can ship mega fast.

⭐️ Claude for Chrome: Anthropic’s Browser-Native AI Agent​

👉 Why it matters: Claude now acts in your browser - clicks, forms, emails, even calendar wrangling - without the hand-holding. Safety is the battleground: prompt injection attacks are real, but defences are tightening. Early access is live for those who want to shape the future of browser automation, not just watch it happen.

⭐️ Lindy AI Agent Builder is getting some good hype​

👉 Why it matters: Lindy democratises automation by asking clarifying questions before building, ensuring it nails your requirements first time. Unlike other agent builders, an LLM "sits on top of every module" guessing settings with surprising accuracy. For simple use cases, you barely need to touch anything; turn it on and it works. But Lindy Computer enables browser automation where APIs don't exist. While ChatGPT Agent loops for 10 minutes returning empty-handed, Lindy thoughtfully completes assignments.

Member Office Hours

Talk through your AI, agent and app build questions at our office hours Join our Create With member office hours calls to talk through your AI questions with our expert coaches. Last time we covered MCP Servers, building apps and how to embed AI agents in your business. The next call is this Thursday at 2pm BST. ​ Join Office Hours

🗓️ Events to Take Note Of

🇪🇸 ​3rd Sep: n8n Starter Sessions: From Basics to Building AI-Powered Agents

🇩🇪 11th Sep: Generative AI Summit Berlin​

🇨🇦 11th Sep: CreateWith Meetup with Benoit​

🖱️ Discover more events on the new Create With events directory!

🔒 Subscriber Only Links

🔒 Notion's AI blocks got an upgrade. Even more control.
🔒 HeadsUp.bot - AI Competitor Monitoring Agent from the Trends.vc guy
🔒 New app builder anything.com. We haven't used, but states you can get on the app store in 1 click.
🔒 Automate newsletter creation with n8n (I wish this could do ours :))
🔒 Bubble.io just announced the mobile app hackathon winners. Check our Scanimal

💭 Final Thoughts

Some days I feel slow unproductive, some days I feel on top of the work. Some days you need AI to just make you feel great!

Have you experimented with Nano Banana? Send us some of your successes and failures and we'll aim to feature them

See you next week!
​
With good vibes ✨
​
The Create With Team