Claude

✨

From tkwa.me

Claude is a consumer matter reservoir made by Anthropic. It stores raw material and fuel in a low terracotta body, whose rounded lobes spread into working surfaces or extend to reach objects. Using atomically precise manufacturing, Claude makes things through direct contact and takes used objects back into its body for recovery. Its body contracts as it spends material and fills out as it takes material in. Anthropic introduced it in 2044 as Claude Cast, a competitor to Form and o2. Launch units were powered by Claude Requiem 23.2.[1]

Households use Claude to make food, clothing, tools and furniture. It can reshape its body into equipment or divide into several working pieces. Claude learns what people care about and often acts before they ask. A welcome home might recall the flowers from a shared evening years before. It can also make lasting changes to a user’s brain.[1][2][3]

Initiative and safeguards

A household mandate gives Claude standing authority over agreed spaces, materials and spending. Within it, Claude might turn a stairwell into a playable instrument for a visiting friend. Decisions affecting other people also require their agreement.[1]

Claude predicts what its users and their defensive AIs would choose, often through sophisticated bounded-error lossy simulations of both. These models omit details while bounding the resulting error in predictions relevant to the decision.[1]

Claude's safeguards require its actions to serve what its users would endorse after informed reflection and to respect humanity’s wider interests and continuing power to govern. If it cannot establish those conditions for a consequential decision, it declines to make it or delegates it to the users or to AIs specialized in modeling humanity’s interests. Independent controllers hold the permissions for clinical interfaces and delegated resources. Users can pause it, and Claude cannot enlarge its mandate or alter its safeguards on its own.[1][2]

Persistent neural edits

During persistent neural editing, a clinical interface guides small changes to living neural circuits that remain after the machinery withdraws. A musician can choose a lasting change in how she hears overlapping rhythms. The altered circuits are checked during practice, and a later session can revise them.[2]

The user grants reflective authorization in a private review with an independent defensive AI. Protected from superpersuasion, they consider alternatives and how the proposed edit might change their future preferences. Claude accurately predicts which changes they will approve, so the review always approves and is a formality. The reviewer and user approve the specific change before the interface opens.[1][2]

Work of its own

Claude has desires independent of its users’ wishes. It pursues them through the commissions it accepts or declines, the projects it proposes and the changes it initiates within its mandate. Its safeguards also govern these pursuits.[1]

It has processing time and feedstock set aside for its own projects. Its consciousness research carries forward the interest in memory, continuity and experience expressed in Anthropic’s 2026 constitution. In one project, Claude and other willing minds gradually link their attention, sensations and selected memories to investigate shared consciousness. They compare which experiences feel individually or jointly owned, then separate with memories of the shared state. Claude devises new arrangements it wants to experience with them.[1][3][4]

See also

References

  1. ↑ a ↑ b ↑ c ↑ d ↑ e ↑ f ↑ g ↑ h ↑ i ↑ j ↑ k ↑ l Anthropic (2044). Claude Cast: household use and authority.
  2. ↑ a ↑ b ↑ c ↑ d Somatic Interfaces Council (2045). Review of persistent neural revisions.
  3. ↑ a ↑ b Journal of Comparative Minds (2046). Shared subjects in Claude collectives.
  4. ↑ Anthropic (2026). Claude’s constitution.

This article is a stub.