ai & agentic product studio

Clarity for humans. Clarity for agents.

An AI & agentic product studio. We design and build products for two users at once — the person, and the agent acting on their behalf. Local-first. Private by architecture.

See the work

Every screen here has two readings. Flip the lens.

scroll

the thesis, made literal

Most software is built for one reader. We build for two.

A person sees a rendered interface. An agent reads structure — roles, names, and the actions it's allowed to take. Same element, two readings.

how a person sees it

how an agent reads it

button {
  name:   "Summarize this"
  role:   button
  action: summarize({ url })
  state:  enabled
}

It's the same button. Clarity for humans. Clarity for agents.

a named discipline

Agent Experience. We call it AX.

UX is how a product feels to a person. DX is how it feels to a developer. AX is how it feels to an agent acting on your behalf — and almost no one designs for it on purpose. We do.

environmental ax — the scorecard

  • NameableEvery surface has a clear, stable name.
  • ReachableThe agent can get to it without guessing.
  • PredictableThe same action does the same thing.
  • ObservableThe agent can tell what happened.
  • ReversibleMistakes can be undone, safely.
  • BoundedIt can act — only within limits you set.

Six principles for designing the environment an agent works in. Drawn from building Clarity — written up in Field Notes.

our specialty

We specialize in local, private agentic systems.

Local-first, private-by-architecture is what we do best — models and data running on hardware you own. It's a specialty, not a cage: when a project genuinely needs cloud models or integrations, we still architect them so your data is scoped, minimized, and as secured as the system allows.

  • Local-first by default
  • Cloud integrations when they earn it
  • Scoped, encrypted data flows ax
  • Private by architecture

Private by default; cloud when it helps — scoped, encrypted, and never by surprise.

data that leaves on its own: 0

the work

Three products. One thesis.

Real, opinionated AI products — evidence of how the studio builds, not a download funnel. Each one is built for both readers. Flip the lens to read them as an agent would.

  1. 01 / 03 in development

    Clarity Browser

    An agent-native browser — private by default, with the user experience still first. It reads each page as a clean accessibility tree and acts through named tools, so you and the agents you delegate to share one calm surface.

    View product Download
  2. 02 / 03 in development

    Mnemos

    A self-hosted personal memory OS — processed locally, never sent to the cloud. Capture from your devices, build a private, linked memory you own, and let agents recall from it through scoped tools.

    View product Download
  3. 03 / 03 coming soon

    OHSOM

    A private, local-first agent network for your whole life. A mesh of agents grounded in your own memory and rules — acting across your life without surrendering it to the cloud.

    View product Get notified

field notes

Notes from building agent-native products.

Writing on agentic development and Agent Experience — what I learn from the inside, while building. The native blog, and the odd Medium cross-post.

about — in the first person

I'm Ahmed Kamal Eldin. I design and build AI products — and the agents that use them.

Clarity is my studio. I work end to end: the vision, the architecture, the interface, and the agent-facing surfaces underneath it. I build products that are private by default, run close to you, and treat the agent as a first-class user — not an afterthought.

If you're building something in this space — or want something built — I'd like to hear about it.