Pane App

You, your AI, and your work. Under one roof.

[Pane App @paneaiapp]

You, your AI, and your work. Under one roof.

Pane: local-first. Notes, tasks, projects, and your AI partner on your machine, under your control.

Bring your own AI: Venice, Anthropic, OpenRouter, Ollama. Pairs natively with OpenClaw.

Mac + Linux: https://t.co/2eYd0Lwq1t

[Pane App @paneaiapp]

Everything you can do in Pane, your AI can do too.

For example, here's your AI, inside the note you're writing.

[Pane App @paneaiapp]

You installed Pane. Now what?

The first things to set up, in order. Five now saves you hours later.

[Matt Carano @matthewcarano]

The AI didn't make your app janky. Your process did. I'd know: I can't code, and my app has paying customers. https://t.co/Cp0Hdr5QFr

[Matt Carano @matthewcarano]

Grok Bot looks genuinely impressive. That's not the issue.
The issue is where it lives.

It runs on their computer, not yours. It signs in with your credentials.

Everything it learns about you, your projects, your habits, your context, sits on their servers under their terms.

Why does that matter? Because an agent gets valuable the way an employee does.

Time. Context. Trust.

Months of accumulated understanding of how you work.

And you can't take any of it with you. Not to another model. Not to another app. Not to your own machine.

Prefer a different environment? Too bad. The agent stays where the company keeps it.

So they hold every lever. They can raise the price, swap the model, rewrite the terms, or shut it down. Your only vote is your subscription.

That's what sovereignty means in practice. Not a buzzword. Two questions:

Is your data on your device or theirs? Can you leave with what you built, or not?

If the answer is no, you don't have an agent. They have a user.

It's why I built @paneaiapp the other way. Your files on your machine.

Bring your own model. Leave whenever you want, and take everything with you.

[Matt Carano @matthewcarano]

I can't tell you how surreal this is. Seven months ago I learned about @openclaw, bought a Mac Studio and started learning.

Six months ago I began building @paneaiapp to replace the need for copy/pasting between Discord (comms), Evernote (documents), and Trello (project management).

Today I can finally say Pane works better with OpenClaw than those three apps combined, because Pane now allows your Agent to reply multiple times without being asked.

It was the last piece in the puzzle to make Discord obsolete.

So I can give Brit a multi-step task, and she completes it, checking in with me after every step.

This is just one of many things that makes Pane special. If you use OpenClaw there is no better interface than Pane.

[Matt Carano @matthewcarano]

YOUR AI DOESN'T KNOW YOU. IT NEVER WILL.

Every time you open ChatGPT, you're talking to a stranger.

It doesn't know you. It doesn't know your work. Tomorrow it won't remember today.

That's not the future of AI. That's the demo.

Here's what's coming: agents that stay with you. For years. Maybe decades. Agents that know your projects, your history, your values. Agents working on your life's missions with you, day after day, compounding.

But understand the distinction. A model is not an agent.

The model is the reasoning engine. Rented, swappable, upgraded every quarter.

The agent is what you build on top: the memory, the identity, the context, everything you've taught it. The model is what it can think. The agent is what it knows about you.

And that changes the requirements.

An agent you'll keep for ten years can't live inside someone else's app. One policy change, one deprecation, one "we've updated our terms" and your decade of teaching resets to zero.

So it has to live on a machine you control. Behind an interface you control. Because agents you control are agents you can teach.

And teaching compounds. Renting doesn't.

This is why I built Pane: the V1 interface for human/AI partnership. Create local agents. Or bring yours from OpenClaw. Either way, your agent stays with you.

Ten years from now, will your AI know you?

Or will you still be introducing yourself?

[Matt Carano @matthewcarano]

In the comments people are complaining about having to spend $300 per month on Grok Bot.

If you want to build something production-grade that a customer will actually pay for, you're going to spend money. I've spent roughly $10,000 in API costs building @paneaiapp.

Before AI, that same product would have cost me hundreds of thousands. Designer. Front-end dev. Back-end dev. Security review. Months of hiring and managing before a single customer saw anything.

$10K to replace all of that isn't expensive. It's the biggest discount in the history of software.

At no point has it been more achievable to launch a real product, even with no dev experience.

The people complaining about $300 a month aren't being priced out of building. They're meeting the price of admission for the first time, and it's never been lower.

[Matt Carano @matthewcarano]

Building with AI is amazing. I've always had ideas I could not execute because I can't code. BUT!

If you are a non-dev building an app with AI you will make janky shit unless you overcome AI's memory issues and context window limitations with systems and procedures.

It took me 6 months to build @paneaiapp but half of that was creating the infrastructure that ensures my code is production-worthy.

You can make whatever you want — the future is now. But to build it right takes discipline.

[Matt Carano @matthewcarano]

Models are getting dumber - been living this while building with AI agents. As models trade stored knowledge for reasoning, they're only as right as the context you hand them.

Too little and they confidently wrong you. Too much and the window floods and they lose the plot.

It's why I built @paneaiapp the way I did: you and your AI work in the same project, so context accumulates where the work happens.

A memory system stores what matters and the AI searches it at will, pulling in exactly what it needs, when it needs it.

You're not prompting a model. You're building a knowledge base you both work in.

Reasoning is the engine. Context is the fuel line.

[Matt Carano @matthewcarano]

Pane (@paneaiapp) has an official OpenClaw skill on ClawHub 🦞

If you run OpenClaw, your agent can now operate Pane directly:

The skill teaches your agent to create notes, tasks, and projects in Pane through the Gateway, manage chat sessions, and sync identity files.

Passed ClawHub's security audit, MIT-0, and the full SKILL.md is readable before you install — as it should be.

Skill page: https://t.co/CvFl15ouf9

[Matt Carano @matthewcarano]

THANK YOU VERY MUCH

It's likely @paneaiapp would not have come to be without these people. Most of them don't know they inspired me or even that I exist.

@steipete - Creator of OpenClaw. A year ago, someone with no engineering background could not have built what Pane is today. OpenClaw changed that. It's one of the most important tools ever made for human-AI collaboration.

@AlexFinn - I saw him posting about OpenClaw back in January. That's what got me to look. One post changed the trajectory.

@BrianRoemmele - Relentlessly optimistic about what technology can do for people. Subscribe to Read Multiplex, like I do. In a world full of AI influencers who cannot articulate the vast benefits of AI, Brian is signal through the noise.

@BowTiedBull - Drilled into my head that unless I build my own thing, I'm a line item on someone else's spreadsheet. Has the most useful newsletter on the planet.

@n3xco - For helping get Pane to the finish line, and for working on what I believe is the biggest problem in technology: AI sovereignty.

[Matt Carano @matthewcarano]

Pane walkthrough, today 12:30 PM ET.

Your AI, with memory that persists. Your work, in one place it can actually reach. Local-first, bring your own model.

Come see it live. https://t.co/avJW7lejxR

[Matt Carano @matthewcarano]

You, your AI, and your work. Under one roof.

Pane is live. Local-first. Bring your own AI: Anthropic, OpenRouter, Ollama. Pairs natively with OpenClaw.

Mac and Linux → https://t.co/qxG4KSqzrg https://t.co/N006eCtk83

[Matt Carano @matthewcarano]

The loudest voices in AI can't answer the simplest question: what do regular people get?

Kevin O'Leary is spending $100 billion on AI.

When asked/He answered:

- What do regular people get? "Nobody knows."
- Why build it? Beat China.
- Why subsidize it? "Welcome to America, buddy."

These questions have real answers. The men spending the billions just don't have them.

And the answers decide whether our AI future is the utopia or the dystopia.