I want Noema to become a place where an idea can turn into something useful.
A clearer piece of writing. An answer you can check. A small tool that solves a problem you keep running into. Eventually, a workspace where you can research, create and hand over parts of a task while staying in control of the work.
That is the direction behind the next chapter of Noema.
As AI becomes more capable, the experience around it matters more. People need somewhere to bring their context, develop an idea and keep working on the result. They also need to understand what they are sharing, where the work happens and what it costs.
I think those things should be part of a good product from the start.
Our direction is simple: a private AI workspace, built to put you in control.
For me, control has a practical meaning. You should have a say in where your AI runs. You should be able to keep conversations on your own device. You should understand when a task uses remote compute, and have useful ways to protect the information you send. The cost should be clear enough that you can make a decision before you commit to it.
Those choices should be understandable in normal language. You should be able to sit down with a task and get on with it.
Noema already brings several parts of that experience together. You can run supported models in your browser, use a model running on your own computer, or access cloud models through Noema. Your conversation history is stored in your browser. Local AI can be used without buying cloud credits, while remote compute is available through wallet-funded credits without a personal signup or KYC.
Local AI matters because some work can be done using the hardware you already have. A draft, a set of notes or a piece of code can be processed there. On supported desktop and mobile browsers, that also means being able to start from a website. The model and the device still need to be a good fit for the task.
Cloud access gives you another option when you want a different model or more capability. ZKR helps protect that path by replacing detected sensitive details before a cloud request and restoring matching values in the reply. Supported documents and receipt images can be reviewed before sending a protected version. Auto routing uses zero-retention cloud routes. The protected request is still processed remotely, so local and cloud remain meaningful choices.
Being able to use these tools without building another personal profile is part of the same idea. Privacy should show up in the decisions the product lets you make, and in what happens to your work.
The next release brings that direction into a more complete workspace. We have been refining the interface across desktop and mobile, improving how people find their settings and models, and bringing generated code closer to something they can actually try through in-app previews. The website and guides are being rebuilt around the same clearer story.
What matters to me is how those pieces work together. Someone asking for a small webpage should be able to see what it does, change it and take the result with them. Someone reviewing a document should be able to understand the answer and check the relevant details. Someone writing should be able to stay focused on the thought they are trying to express.
That gives us a useful standard for deciding what to build: does this help someone finish the thing they came to do, and does it give them more control along the way?
The audience is broad. A student working through a difficult topic, someone drafting a proposal, a developer making a prototype and a person reviewing a crypto or a financial transaction can all use the same underlying tools. Their tasks are different. They should be free to decide what Noema becomes useful for.
From here, the plan grows into multimodal capabilities, deeper research, connected knowledge and agents that can help carry out work with clear permissions. We also want developers to be able to build with Noema, and eventually extend the platform with more independent compute. Those are broader milestones we will work toward and release in useful steps.
I want each addition to feel considered. Easy to understand, useful on its own, and better because it belongs in the same workspace.
In the coming releases, we will show that progress through real tasks: what someone starts with, what Noema helps them do, and what they can take away. Alongside that, we will explain the choices behind the product in plain language.
The goal is a place you can think, write and build with AI, with your work and your choices at the centre of the experience.
That is the Noema we are building.
