We use the term 'account' to refer to both users and organizations (aka. "orgs") in HASH.

All users and orgs have an account name, such as @example. This can be used to access an account's public profile, such as at hash.ai/@example.

Every user and every org has their own web, containing entities (data) and types (descriptions of data).

User webs are known as personal webs, and org webs as shared webs.

You can access any number of shared webs, in addition to your own personal web.

Data in a web exists in the form of entities. Entities have:

• →

  An entity type which describes “what they are” and what kinds of attributes may be associated with them

• →

  Any number of attribute values which contain an entity’s properties and links (structured data)

• →

  Any number of claims (observations about an entity that don’t fit into its existing/available attributes)

• →

  Any number of notes, which allow completely unstructured information to be associated with an entity

An individual entity may have one or more entity types, which describe what sort of thing(s) an entity is, and the kinds of structured information (known as attributes) that can be associated with it.

Attributes are structured information stored on an entity’s record. Their values are either properties (which contain data) or links (which point to other entities).

Every link attribute has a link type which describes what its value refers to and provides it with a label.

Constraints on the specific entities, or type(s) of entities that a link may point to, are also defined by the link’s type.

Example link type: Employed By (may point to Company)

Every property attribute has a property type which describes what its value refers to, and outlines the format(s) of data (i.e. the data types) that it accepts.

Example property type: Average Temperature

While property types provide semantic meaning to property values, data types define the range of accepted values, and specify the labels that should be shown in conjunction with them.

Example data types: Celsius and Fahrenheit

History in HASH refers to three things:

1. →

   change tracking: a complete record of all the changes to an entity or type over time

2. →

   versioning: the ability to inspect, access and even use older versions of an entity or type

3. →

   provenance: the ability to see where information and changes originated, as well as what caused them

HASH agents are AI agents which work for you. They perform tasks on your behalf, helping grow your web with useful information, and keep it up to date.

You can also ask your agent questions about the data in your web, helping you find and use relevant information more effectively.

Agents can be given goals. When provided with a goal, available agents will analyze your aims, develop a plan, and execute it autonomously upon your behalf, much as a capable human colleague might.

Right now, HASH supports research and analysis goals, helping you capture and process information about the world.

Agents can also reliably and automatically execute sequences of steps, known as flows. In contrast to goals, the actions involved in a flow are predefined ahead of time.

Flows can be triggered on-demand, in response to observed events, or on a recurring schedule.

Experiments allow you to run simulations of the real world safely and inexpensively in-silico. Running a simulation helps you explore "what if" scenarios, identify optimization opportunities and risks, and stress-test decisions before acting on them.

Your inbox contains notifications, messages, and actions (including drafts).

For example, if an agent requires you to clarify some part of a stated goal, a question from it will appear here.

Plugins allow you to interact with your web within other applications, and typically provide HASH with additional capabilities.

For example, installing the HASH browser extension lets you (optionally) allow your agents to browse websites using any accounts you may be logged into. This can let them bypass paywalls, and access other protected content, exclusively on your behalf.

Integrations enable HASH to act on your behalf in external applications, and make it easy to sync data with your web.

Existing data sources can be connected to HASH, and the information within them represented as entities in your web, which are kept up-to-date in realtime.

Every entity and type has a set of permissions you can control, which govern what other users can both see and do.

Permissions can be managed very granularly, allowing view, update, comment, share and other forms of access to be restricted only to authorized users.

Sharing allows you to make select entities, types, or portions of your web available to others. You can also jointly manage shared webs with others, through organizations.

We're constantly expanding what HASH can do. Upcoming capabilities — including pages, blocks, and apps — are outlined in our longer-term vision.