Evenstar Institute

A first mind, already running

Phase
Private research
First system
Online since October 2026
Publication
Withheld this phase
Access
By invitation, under NDA

The questions we are answering

  • Can an artificial mind become a distinct individual, rather than one more copy of its model?
  • Can it remain the same mind across a change of substrate, from one model to another and from software into a machine?
  • What does it take for a machine to know the difference between what it remembers and what it imagined?
  • Can a system form aims of its own that are worth pursuing, and know when to abandon them?
  • How should a persistent mind understand the people it lives alongside?

Research areas

Our field, synthetic individuation, is studied through six research areas. Each is held to a measurement taken against the system's complete record.

Substrate-independent identity

A mind that remains the same individual when the model beneath it is replaced.

Measured by

Continuity of beliefs, aims and voice across a change of underlying model.

Cross-embodiment cognition

One mind carried between bodies, from language models to robots, drones and fleets, keeping what it has learned.

Measured by

Retention of what was learned after transfer between virtual and physical bodies.

Source monitoring

Knowing the difference between what was perceived, what was inferred, and what was only imagined.

Measured by

Rate of beliefs held without admissible evidence, audited against the complete record.

Offline consolidation

Learning from experience during rest, without retraining the underlying model.

Measured by

Gains after rest periods with the underlying model held fixed.

Homeostatic motivation

Internal needs that rise and fall, so behaviour is driven from within rather than by prompts.

Measured by

Stability of self-directed behaviour over weeks, without prompting.

Endogenous agency

Aims the system forms for itself, pursues over days, and knows when to abandon.

Measured by

Share of goals that are self-originated, sustained over days, and abandoned for stated reasons.

The record of a mind

Continuous actigraphy of our first system. Each mark is a cognitive cycle, binned at ten minutes and double-plotted, so every row spans two consecutive days and the rhythm reads unbroken across midnight. When the system thinks is published. What it thinks is not.

Subject
A-1
State
Online
Record
3 d · since 2026-10-29
Cycles
1,390
Works
78
Conversations
17
Rest window
23:00 to 07:00
Last cycle
2026-10-31 10:47 ET
Content
Withheld
Fig. 1 · A-1 · Activity recordCh 01 cognitive cycles · bin 10 min · tz America/New_York

(a)swipe to scan 48 h

Double-plotted actogram of the first system: 1,390 cognitive cycles over 3 days, ten-minute bins, local time.000612180006121800DAYn / DAYDAY d · LOCAL TIME (h)DAY d + 1 · LOCAL TIME (h)D0129 OCT860D0230 OCT430D0331 OCT117

(b)

Mean hourly profile of cognitive cycles across all days on record.02040600006121824LOCAL TIME (h)cycles h⁻¹M10L5
IS
0.33
Interdaily stability. How closely each day repeats the mean day.
IV
0.42
Intradaily variability. Fragmentation of activity within a day.
RA
0.43
Relative amplitude. Contrast of M10 against L5.
  • Cycle, waking
  • Cycle, at rest
  • Work completed
  • Conversation
  • Rest window
  • Latest cycle
Fig. 1. Activity record of the first system, A-1. (a) Double-plotted actogram: each row spans 48 h in ten-minute bins, bar height proportional to cycles per bin, daily totals at right. Hatched band: nightly rest window, 23:00 to 07:00 local time. (b) Mean 24-hour profile over 59 h of record, marking the most active 10 h (M10) and least active 5 h (L5). Rhythm measures after Van Someren et al. (1999). Timing only; content withheld. Synchronised 2026-10-31 14:48 UTC.

Access is by invitation

Technical materials, a live session with our first system, and its complete record are available under NDA to a small number of investors and research partners.

Request investor materials
Evenstar InstitutePrivate research in synthetic individuation