Hello and happy Tuesday! We have big news!

Our flagship conference is coming back in 2026 and early-bird tickets are on sale NOW.

Let’s jump in…

🔎 Preview: In today's issue

• 🤔 Agentic workflows inside Notion
• 📺 From coffee chaos to automation pro
• 🐒 AI in the wild: real‑world uses

🚨 BIG NEWS... Create With 2026 tickets are on sale...

...and we're moving the conference to the beach! Brighton, UK to be precise.

And it's not just the location that's changing. We want this conference to be the most action-oriented, hands-on learning experience possible.

Our focus is practical, real-world AI and we want everyone to leave having built something (or at least extremely confident they can build something).

What to expect

💡 Keynotes + case studies from teams already using AI in production

🛠️ Hands-on sessions with no-code tools, workflows and agents

🤝 Partner talks + live demos showcasing what’s possible right now

We moved to a bigger venue on the Brighton seafront 🌊 and lowered our prices to welcome more builders, founders, operators and product teams than ever.

​🎟️ Grab your early-bird ticket for just £35 (limited time only).​

Matt got bored writing the newsletter so wanted to see what Brighton looks like on a stormy day with Gemini Nano Banana Pro

🤓 Explain Like I'm 5
​

For those who don't know, Notion is a super powerful all-in-one workspace that lets you create, organise, and connect documents, databases, tasks, and knowledge in a highly flexible, Lego-like system.

A few months ago Notion introduced an AI Agents feature, where you can create multiple agents that follow their own custom instructions and can access all your Notion data.

Create With team member Matt quickly realised he could effectively build his own custom software simply by storing and referencing various procedures and instructions.

Here's how he did it

The mental shift: Stop thinking of AI as a chatbot that answers questions. Start thinking of it as a team of specialists you can delegate to.

Instead of "help me write an email", you build a prompt that acts like a job function. A copywriter. A research assistant. A data analyst. Each with specific instructions, tone, and output format.

This works across any LLM: Claude, ChatGPT, Gemini, Notion. The principle is the same.

Why it matters:

• Consistent outputs (almost) every time*
• Complex tasks broken into repeatable steps
• You write instructions once, reuse forever

*It's not bulletproof yet but in our opinion if you have to do 10% of the task in 10 minutes rather than it taking 2 hours...it's a win.

The kicker: You're essentially building software, but in plain English. No code. No visual canvas. No connecting boxes with arrows. Just describe what you want, step by step.

📺 Don't want to read? Here's Matt explaining it

How it's set up

For each of his agents, Matt sets up a custom instructions page, a permanent document the Agent references every time. Update it once, every future output reflects the change. This is basically dummy proof and you can set this up in minutes.

This is the same concept as ChatGPT custom instructions

Unfortunately Notion doesn't allow for a sub-agent structure. One Agent = one set of instructions.

The hack Matt built: A Notion database called Prompts. Each row is a sub-agent with a specific job function.

When Matt needs a task done, he tags the relevant prompt using Notion's system of "@ mentions". The Agent pulls those instructions and executes the task.

Inside each prompt: Step-by-step instructions written in English. What data to create. What properties to add. What format to output. Like an automation workflow, but readable.

Here's another example Matt uses for new leads.

When a call with a new lead has ended and wants to store the lead, he "@ mentions" the sub-agent called New Lead, and it creates the necessary records according to its instructions.

❓ What to watch out for

• This agent type experience is only useful inside Notion. You can go get data from Slack, Gmail, the web and other sources but you cannot trigger actions outside of Notion. At least not right now.
• Notion Agents are only available in the business plan

💡 Tip: Start with one sub-agent. Get the output right. Then duplicate and adapt for the next job function.

P.S. Custom agents are apparently coming to Notion soon and could effectively replace the need for the prompts hack above.

​

Upcoming Create With AI Meetups

📅 Create With: Norwich AI Meetup 📅 Mon 12 Jan 2026  · ⏰ 18:00  · 📍 Norwich

📺 From Coffee Entrepreneur to Automations Expert

​Tiberiu Socaci tells the fascinating story of how automating his chaotic coffee-subscription business pulled him deep into Make and n8n, eventually leading him to launch Make It Future, a fast-moving automation and AI-agents studio.

He shows how the same hacks that saved his coffee business now power full apps and enterprise workflows across Europe. Along the way he breaks down how to set expectations, test AI properly, and pick the right mix of tools like Bubble, WeWeb, Bolt, Lovable, Supabase and Xano.

He also dives into some pro-level insights on building efficiently and effectively with AI.

Thank you for sharing Tibi!

▶️ Watch Video​

💡 Want to share your product with 6000 readers?

​Click here to help us to keep the lights on and become one of our sponsors. We have a consistent CTR of over 40% and an audience of tinkerers, builders and pros.

🔊 On your radar

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

⭐ Stripe’s Black Friday city: real‑time tribute to small businesses​

👉 Why it matters: Stripe built a miniature city that shows live activity from shops using Stripe over Black Friday and Cyber Monday. It’s a playful dashboard that turns payments data into something you can see. For non‑technical teams, it’s a neat example of turning backend events into a simple, visual story your audience can grasp. This stuff is only possible with generative AI as it would costs too much to produce this UI otherwise. Could you do the same?

⭐️ Claude now auto-compacts earlier chat context to avoid hitting limits​

👉 Why it matters: Long chats often break when you hit the context cap. Claude now trims and summarises older parts of the conversation on its own, so you can keep working without starting a new thread. For workflows and AI assistants, this means fewer lost chats, smoother multi-step tasks, and less manual pruning of prompts. There are still reports of mixed results with images and long coding threads, so expect improvements over time but try it for day-to-day chat continuity.

⭐️ 10 beautiful websites you can build this weekend with Replit and Gemini 3​

👉 Why it matters: These are ready-to-try examples with prompts and live demos, ideal for builders like us. Pick one, copy the prompt, and ship a small project fast. Useful for portfolios, product pages, simple games, and creative sites. If something breaks, simplify assets and keep scope tight to get a working MVP. We've been amazed by the quality of front end sites produced by Google's Gemini Pro 3.

😬 Extra bytes

⭐ Lovable has a mission site to learn to build real apps fast.
⭐ OpenAI for Startups: credits, events, and build support​
⭐️ There's still loads of Black Friday offers currated by Tony
​⭐️ n8n has released MCP support for local instances (so your lovable app can talk to your workflows)
​⭐️ Airia hackathon in UK, December 15th with cash prizes​

PS...

If you want to learn more about automations we've partnered with 100 School's Agent Bootcamp. All Create With members get 20% off all bootcamps.

💭 Final Thoughts

...Actual image of a vibe coder?

See you next week! And don't forget to vote for the Bubble.io yearly community awards
​
With good vibes ✨
​
The Create With Team