Interactive preview available · Sign-up is not open yet.

An AI workspace for people and teams.

Your thinking partner.
Your private workspace.

Think things through with AI that learns how you work. Carry your knowledge between conversations, delegate on your terms, and keep control of what stays private.

Explore your thinking spaceTry it out

No account needed Interactive preview with sample data

A better conversation · illustrative
Help me think this through before we act.

Let’s get on the same page.

Where do you want
to stay involved?

The designThe implementationThe decisions
YOUR CALL

“Work through the options with me.
Then help me carry out the plan.”

Direction agreed ✓

Discuss the direction → Make a plan → Move it forward

01 / A partner in the thinking

Think it through together.
Then take it forward.

Some days you want to talk through a difficult decision. Other days you know the direction and want help carrying it out. A good AI partner should make room for both.

You choose where to contribute—design, implementation, the details, or the final judgment. We’re building an assistant that learns your preferences, discusses the important choices, and takes on the work you choose to hand over.

  • A discussion that helps you find your direction.
  • An approach you understand before work begins.
  • Execution with room for your judgment.
Find your thinking space ↗Why we’re building with people in mind ↗

02 / Context & memory

Know what carries forward.
See what still needs you.

What did we decide? Which notes matter here? What is still open? A continuing conversation needs more than a long chat history.

01 / Remember

What matters to you.

Notes, preferences, and the decisions worth carrying forward.

How I like to work Discuss the options before choosing.
A decision we made Keep the first release focused.
02 / Focus

What matters now.

Bring selected references into the task. Keep the purpose in view.

↗ Project brief
◇ Notes you selected
◎ The current goal
03 / Continue

A next step you can see.

Follow the work and find the questions that need your judgment.

Needs your inputWhich direction feels right?

We give saved knowledge, selected references, and work in progress their own place. Keeping the right context close helps the AI stay focused—and helps you return to the work without reconstructing the whole conversation.Illustrative context views · product in development

Inside context and memory ↗

03 / Privacy by design

Your knowledge helps you.
We respect its privacy.

Your experience, unfinished thoughts, and way of working are valuable. Your AI uses that context to help you. Our design keeps it out of routine staff review, model training, and conversation-text analytics.

  • Your content is not training material.
  • Managed model calls use approved zero-data-retention routes.
  • Private and confidential information has explicit sharing boundaries.

AI processing uses selected content. Limited human-access exceptions, company settings, and connected-tool terms are explained in our privacy guide.

See how your privacy works ↗

04 / Your AI command center

Your agents, coordinated.
Your context, under control.

You may already work with several AI agents. We’re building a place to coordinate their tasks, carry the relevant context, and see the progress and decisions that need you.

Start with the local integration work for Claude Code and Codex. Choose which folders and items they may use, while keeping private material outside that scope.

Integrations are in development. Use your own eligible accounts; each provider’s terms, limits, and privacy rules apply.

Your tools, your sharing choices ↗
Local integrations · in development
Bubbling AIYou set the direction.
⌘

Claude Code

Your eligible account.
Your selected folders.

↗

Codex

Your eligible account.
Your selected folders.

ContextProgressYour decisions

Specialist tools. One place to keep the thread.

A working approach · illustrative
  1. 01
    You choose the destination.

    Start with your goal and what matters.

  2. 02
    Agree on the route.

    Discuss the approach and key decisions.

  3. 03
    Give the work room to move.

    Useful execution within the agreed boundaries.

  4. 04
    Come back to your judgment.

    Review, refine, and choose what comes next.

05 / Built to evolve

Built to work with
better models.

Your workspace should benefit when models improve. We keep adapting the workflow so yesterday’s workaround doesn’t become tomorrow’s obstacle.

Focused context supports both useful answers and efficient use of AI. Different model strengths can serve different parts of the work. Important permissions and workflow steps stay enforced by the system, while we revisit the scaffolding around the model as its abilities change.

How we work with the model ↗

The thinking behind the product

A future with
you in the picture.

Read the journal ↗

Get to know your workspace

A space to think.
A partner to work with.

Explore the conversations, memory, and work surfaces today. When live signup opens, continue with your email for a free week of included AI usage.

Try it out ↗Preview email signup

Today’s preview uses sample data and simulated replies. It does not create an account or activate a trial.

A few things to know

Good questions.
Clear answers.

What is Bubbling built for?

A continuing partnership with AI: thinking together, remembering useful context, and carrying out work with your judgment involved. The workspace brings conversations, memory, files, artifacts, and work state together around that relationship.

What can my company see?

Our company-workspace design separates private thinking from shared work. Personal to each member protects private chats and work memory from ordinary company-admin access. Held by the company permits company access. Shared project conversations, documents, and uploads follow company-workspace rules. The selected policy and exceptional-access rules must be clear before you join.

Does private mean everything runs on my device?

No. The hosted service needs to process selected content, including through approved model and automated safety-check routes. Our direction is no training on your content, approved zero-data-retention model routes, and no routine staff reading of private conversations. The privacy guide explains the specific boundaries and limited access exceptions.

Can I keep using Claude Code or Codex?

That is the direction of our local integration work: bring context, tasks, and progress together around tools you already use. These integrations are still in development. Your own eligible accounts and each provider’s terms apply; outside tools do not inherit our managed-model privacy policy.

What can I try now, and how will the free week work?

You can explore the actual frontend with sample data and simulated replies today. It does not save your work after refresh or create an account. When live signup opens, we plan a free week of included AI usage with your email. Allowances and paid plans will be published before then.