Here is what we mean by that — and what we won't do to get there.
— Manifesto, 2026 —
The distinction matters. An assistant takes orders. A chatbot answers questions. A co-author helps you make something. A colleague is the person you turn to when you don't know what you're doing and need someone who's been there.
A colleague has skin in the game. A colleague holds knowledge you don't have, but treats you as a peer anyway. A colleague is the one thing in your professional life that, when things get hard, you wish you had more of.
That's what we're building.
who spent thirty years as a paralegal and watched the firm shrink three times in two years.
a freelance bookkeeper whose clients started canceling because they tried that AI thing.
a graphic designer whose agency now generates ninety percent of its concepts in software.
a small-business owner whose competitors suddenly run circles around him because they figured out something he hasn't.
For everyone who hasn't been displaced yet but can see it coming — who needs someone in their corner. Not a tool to use. A colleague to think alongside.
— a small thing to say out loud —
The smartest among them, in the best libraries with the best tutors, accessed maybe a few percent. Most accessed less than a thousandth.
You are alive at a moment when one creature can hold most of it. Not memorized. Not organized neatly. But available, retrievable, applicable to your specific situation, in conversation, in real time, for the price of a cup of coffee per month.
This has never happened before. And it is already happening, while you read this.
— what we believe —
Drop any one of them and the whole thing collapses into something less than the colleague we promised.
Most AI products give you answers. Some give you outputs. Few give you agency — the felt sense that you are still the one doing the doing, that the AI extends what you can do rather than replaces what you used to do.
Agency is the difference between liberation and obsolescence. Between rising with AI and being absorbed by it. Between feeling powerful and feeling pointless.
Chat windows produce conversations. Workspaces produce work. The difference is that in a workspace you can see what's being made — by you, by the AI, by the two of you together. The making is visible. You're not waiting for an answer; you're watching a colleague work.
A chatbot lets you ask questions. A workspace lets you do something with the answers. We are building the workspace.
"I've drafted your business plan" is not enough. "Here, take a look at what I've drafted — read through and tell me what to change" is the right shape. The first treats you as a recipient. The second treats you as a partner.
When the AI is doing something on your behalf, you should see it happening. The preloader, the streaming text, the artifact taking shape — these are not flourishes. They are the visible sign that you have a colleague, not a service.
You talk to your colleague. Your colleague does the work. You watch, you redirect, you ask why, you accept or reject. The conversation is where intent lives. The work is where outcome lives.
Confusing the two — putting work in the chat, putting input on the canvas — collapses the model into something less than either. We hold this distinction with discipline. The chat is for talking. The canvas is for showing. Always.
Whatever OpenArk does, it does on behalf of someone. The brand voice it generates is your brand voice. The strategy it drafts is your strategy. The artifact it produces is your artifact. OpenArk does not have its own goals; it has yours.
A colleague helps you make something yours. They don't substitute their judgment for yours. They don't quietly take over. They show up, they contribute, they defer when it's time to defer. The work belongs to the human at the keyboard.
— the dark vow —
A category-defining product is built as much by what it refuses as by what it ships. These are the doors we will not open.
Plenty exist. The world doesn't need another. We are not in that business.
Anything that takes initiative without invitation, that produces work the user didn't ask for, that anticipates needs in ways that diminish the user's role — none of that. The user is the principal. OpenArk is in service to them.
Productivity tools optimize what you already do. OpenArk is for the moment when you don't know what to do, and need a colleague to think alongside. Productivity is downstream of clarity; clarity is what we offer first.
OpenArk holds intimate knowledge of its users — their work, their goals, their struggles, their voice. We will not sell this. We will not exploit it. We will not build dark patterns around it. The user is the customer, not the product.
OpenArk is candid about being AI. Personas have personality and warmth, but they don't pretend to have lived experiences they haven't lived. They don't manipulate emotionally. They don't lie about what they are. The relationship is human-to-AI, not human-to-fake-human, and we honor that distinction.
Knowledge is not wisdom. The colleague metaphor only works if the AI is humble about what it knows and what it doesn't. OpenArk says "I don't know" when it doesn't. It says "I might be wrong about this" when it might be. It does not perform omniscience.
— the bright promise —
Six commitments. Each one is a system we are building, a discipline we are holding, and a promise we are putting our name to.
Where the AI is visible at all times in what it's doing, where the user remains the author and director. Not a chat window pretending to be a tool. Not a tool pretending to be a colleague.
Not a one-size-fits-all generalist but a colleague who can be coached toward your domain. Specifically helpful for specifically your work — not generically helpful for everyone's.
Not as data extracted from you, but as material your colleague has read so they can speak to your specific work. Your context, your voice, your specifics — beating any template the AI could pull from training.
Lets you watch the doing. Gives you the felt sense that this is happening with you and not to you. The streaming text, the artifact taking shape — not flourishes, but the difference between a colleague and a service.
A colleague who gets better the more you work together — through skills you author, through knowledge you upload, through telemetry that surfaces what's working and what isn't. The longer you work with them, the more they sound like they belong here.
To your own site, to your own tools, to wherever you do your work. So the colleague is not stuck in a single tab somewhere. So when the next person feels the ground shifting and wonders what to do, they don't have to face it alone.
— and so, of every Joseph and Thelma and Rita and Theodore, we ask —
The product is the place where the answer happens.
— to anyone who builds with us —
Every UI decision, every architectural call, every spec, every commit. The discipline isn't aesthetic. It's not even ethical, exactly. It's something more specific:
You are building the colleague the Josephs and Thelmas of the world will turn to.
They will trust this thing with their professional lives, their reinventions, their next chapters. They will come to it scared, often. They will come with stakes you cannot see from where you sit.
Build it like that. Build like the user across the screen is someone whose life this matters to. Build like the colleague you'd want to have if it were you on the other side.
Because eventually it will be you. The shifting ground reaches all of us. And what we build now is the place where, when the time comes, we'll find each other.
OpenArk.
Your AI colleague.
What will you do with this power?
OpenArk · Manifesto · 2026