I spent years manually hunting for bubbles. So I built an AI agent to do it for me.

For the last few years, a big part of my work was helping startups grow through organic content.

The actual writing was rarely the hard part.

The hard part was figuring out what was worth writing about before everyone else started talking about it.

My workflow was honestly ridiculous.

I’d open X, Reddit, LinkedIn, Hacker News, Product Hunt, GitHub, Discord communities, and a bunch of random tabs. Then I’d spend hours reading posts, comments, complaints, launches, and small discussions trying to understand where attention was moving.

Sometimes I’d catch something early and the content would take off.

Other times I’d spend half a day researching a topic, publish the post, and realize I was already late.

Eventually I started calling these early conversations Bubbles.

A Bubble is a topic that is still relatively small, but suddenly starts growing across several communities. By the time it appears everywhere on your timeline, the best opportunity is usually gone.

I kept looking for a tool that could do this research for me.

Most trend tools showed broad search data. Social listening tools gave me dashboards with thousands of mentions. AI writing tools could produce a post, but I still had to tell them what to write about.

None of them really answered:

• What is starting to grow right now?

• Why is it growing?

• Is it relevant to my startup?

• Is it still early?

• What should I actually do with it?

So I started building Subio Bubble Agent.

It continuously scans places like X, Reddit, LinkedIn, Hacker News, Product Hunt, and GitHub, then groups related conversations into Bubbles.

For each Bubble, it tries to explain:

• where the discussion started;

• how quickly it is growing;

• which communities are picking it up;

• what people are excited or frustrated about;

• whether the topic is already saturated;

• which content angles are still underused.

Then it can turn that research into an X post, LinkedIn post, Reddit post, thread, launch angle, or content plan based on what you’re building.

But I didn’t want to make another dashboard that people have to remember to open.

The main interface is messaging.

You can text the agent through iMessage and ask:

What Bubbles are growing today?

Why is everyone suddenly talking about this?

Which one is relevant to my startup?

Turn the second one into an X thread.

The agent can also reach out first when it finds something important.

Something like:

A new Bubble around browser agents is accelerating across X and Hacker News. Opportunity score: 91. Want me to explain why?

Eventually, for higher-priority opportunities, it can call you instead of waiting for you to check another notification.

I’m also building an MCP server so people can use the same Bubble data directly inside Claude, Codex, or Cursor.

So while working in Claude, you could ask:

What are today’s biggest developer-tool Bubbles?

Then:

Compare the top two.

Then:

Use the winner to create a launch angle for my product.

Claude handles the conversation, but Bubble Agent provides the live internet context.

The way I think about it is:

iMessage is where the agent reaches you.

MCP is where you reach the agent while already working.

The web app is where you go deeper into the data and configure everything.

It’s still an early version, and I’m building it around the workflow I used manually for years.

I’d genuinely love to understand how other founders and marketers solve this today.

Do you already have a system for finding topics early, or is it still mostly scrolling, bookmarks, group chats, and luck?