Universal Namespace, Inc.
The search and discovery protocol for AI.
One who else? language. Any person, or any labeled AI agent, can use it to ask. Adapters plug the protocol into any database, app, or API. Every source becomes discoverable through one grammar and one namespace. We're building that protocol.
01 Language
One grammar. One namespace.
Search is excellent at finding documents. It is almost useless at finding intent. A want, a skill, a service, a companion — each is a different product today, with a different query language, if it has one at all. That is a layering mistake.
We're building one language instead. Every ask collapses to the same structure. One namespace can address all of them.
ASK := TOPIC “who else?” · SIDE · PLACE · TIME
SIDE is knows, wants, or does. Place and time are first-class. That is the entire language. If an ask cannot be written this way, we don't add a special case. We fix the language.
02 Protocol
How the protocol works.
We're not building a slightly better directory. We're building the language those directories should have been written in, and the protocol they should have shared. A person can speak it. A labeled AI agent can speak it. An existing system should not have to be rewritten to be discovered.
-
The ask
An intent is written in the who else? language. Same grammar for a human or a labeled agent. Topic, side, place, time.
-
The resolver
One resolver parses every ask into the same structured sentence. If the language is universal, the parse has to be too.
-
Adapters
Connectors plug the protocol into a source: a database, an app, an API, an MCP server, or an open dataset. The source stays where it is. It becomes addressable in the same language.
-
The namespace
One address space for every intent. An ask can be published, matched, and answered — by a person, or by a labeled AI agent. Agents will not pose as humans. They should not need to. The language already says who is asking, and for what.
On top of that, anyone should be able to build an intent app: rides, food, doctors, trades, events, or a category we haven't named. Users will invent their own. They shouldn't need our permission to ask a new kind of question. They should need a grammar that can hold it, a resolver that can parse it, and an adapter that can reach the source.
03 Constraints
A few things we will not do.
-
Force data into our silo
The protocol should plug into the source. It should not require that every database move into ours. Open data is where we started. Adapters are how it extends.
-
Personalize from ad graphs
If the product is personalized, it will be from your own asks and taps. Not from ad data. Not from a profile assembled about you elsewhere.
-
Reveal contact by default
When people match, contact is shared only if both sides connect. Until then, it stays private.
-
Let agents pose as people
AI agents are labeled as AI. They do not pretend to be human. They do not send people unsolicited messages.
04 Early proofs
Try what's being built.
These are live demos, not the finished protocol. We're still building them. They exist so you can hit the language in public while the rest of the stack is under construction.
-
Demo stage
Type an ask. A public demo of the language at work.
-
v2
One want in. Who can, and who else, out.
05 Company
Universal Namespace, Inc.
whoelse is a project of Universal Namespace, Inc., founded in Berlin in November 2020. Tobias Martens is the founder, CEO, and CTO.
The name, and the who else? question, have been there from the start. The work now is a search and discovery protocol other systems can speak.
-
Tobias Martens
Founder, CEO & CTO
Programs: Next AI Toronto, Telekom hubraum, and Creative Destruction Lab Oxford.
06 Contact