HomoErectus aims to build a continuing digital individual that generates mental activity from its own goals, desires, emotions, and experiences, chooses action or non-action, and allows behavioral results to keep influencing itself.
Validation needs to ask whether these mechanisms actually participate in the same individual's life. Where did an idea originate, why did it receive attention, how did it lead to action, and how did the result change later mental activity and behavior?
We therefore focus validation on traceable states, experiences, choices, and feedback relationships. This research provides a basis for exploring the formation of consciousness. Whether subjective experience exists requires further theory and evidence.
1. First establish the claims to be validated
The project follows the development sequence “life first, then work.”
Here, being able to live requires an individual to maintain continuing concerns, generate attributable ideas, use real capabilities for meaningful activity, retain experience, and use it later. It also needs to maintain relationships and responsibilities and continue after sleep or interruption.
The individual can depend on models, computing resources, tools, and social collaboration. Autonomy needs to operate within these conditions: how it understands external stimuli, forms its own directions for activity, and makes choices within existing responsibilities, abilities, and boundaries.
Validation therefore concerns whether activity actually occurred and why it occurred as it did. Completing a task can provide evidence of execution capability. Explaining autonomous mental activity also requires tracing the origins of motivation, ideas, and deliberation.
The main project organizes these requirements into life scenarios and uses joint acceptance to examine their connections within the same individual.
2. How seven capabilities are examined together
The first stage requires a self-model, long-term goals, desires, emotions, six-domain memory, external stimuli, and endogenous ideas to be complete and to work together.
| Capability | Validation focus |
|---|---|
| Self-model | Understanding of abilities, limitations, and dispositions is grounded, cautiously revisable, and influences subsequent choices. |
| Long-term goals | Continuing concerns actually participate in attention, learning, and deliberation, with sources and history retained for goal formation and revision. |
| Desires | Satisfaction, gaps, and drive have explainable temporal and experiential changes and participate in idea formation and action assessment. |
| Emotions | Grounded appraisal of experience produces changes that participate in attention, judgment, and later behavior. |
| Six-domain memory | Current cognition, experience, knowledge, work structure, methods, and identity-related content retain their respective roles and can be saved, retrieved, and used. |
| External stimuli | Actual inputs retain provenance and ownership, distinguish feedback, new requests, additional information, and background material, and produce appropriate effects. |
| Endogenous ideas | Without a new direct task, the individual can form ideas from its own state and concerns, deliberate, and undertake activity or reasonably defer it. |
Examining capabilities separately can establish whether local mechanisms are correct. Joint acceptance also needs evidence of connections among them.
For example, an experience that revises the individual's understanding of an ability should matter in a later action choice. An investigation arising from a knowledge gap should produce results that can enter memory and influence later understanding or activity.
The main project's LIFE-21 joint gate requires these connections within the same individual's history. Strong performance on other tasks cannot offset a missing capability.
3. Following evidence through an activity
A useful evidence chain starts before the activity and extends to later use of its results.
Before the activity: what state was the individual in?
Relevant long-term concerns, desire and emotion states, existing understanding, recent experience, current responsibilities, and available capabilities need to be recorded.
These explain why something might draw attention and provide a starting point for checking subsequent changes. Priors supplied at birth need to be distinguished from understanding later formed through experience.
Ideas and deliberation: why consider this, and why choose this way?
The source of an idea and its development into a possible activity need to be recorded.
After attention and decision-making begin, validation must examine what the individual actually used: whether it connected current goals, psychological state, and responsibilities, identified missing grounds, and formed a decision to act, refuse, or defer.
The system can retain inspectable grounds and sources for these judgments without publishing the model's complete raw reasoning.
Action and results: what actually happened in the environment?
Actual operations, corresponding artifacts, returned results, and checking judgments need to be retained.
File contents, command results, message receipts, and external feedback must correspond to the relevant activity. The scope of the conclusion supported by an observation also needs to remain clear: local success supports the corresponding local requirement, and an unknown outcome remains unknown.
Internal change: what did the experience leave behind?
Validation needs to examine how experience entered memory, whether it revised factual understanding, self-understanding, or methods, and whether desires, emotions, and goal progress changed on an evidenced basis.
After saving, independent read-back is needed to confirm that the write occurred and that versions and sources correspond.
Later use: what difference did these changes make?
Finally, evidence is needed in subsequent mental activity and behavior.
For example, did a failure lead the individual to arrange checking earlier? Did new understanding change a judgment? Could it use a learned method in a similar situation and recognize when the method did not apply?
The full relationship can be summarized as:
Actual background and sources → ideas and judgments → action or reasonable non-action → observed results → internal change and durable read-back → subsequent cognitive use and behavior.
Actual records need to connect this relationship. A fluent explanation written afterward remains material requiring further verification.
4. How autonomy and the effects of psychological state are validated
Autonomous activity needs naturally generated sources.
Testers can introduce actual materials, environmental changes, or social interaction and observe how the individual understands them. If a test directly writes the final idea, promotes its intent, or specifies the choice in advance, those interventions need explicit records.
Such controlled conditions can help check whether downstream paths connect. Establishing natural autonomous activity also requires observing how formal mechanisms generate ideas from actual state and experience.
Automatic work continuation or scheduled reminders can establish part of sustained processing capability. Validation also needs activities arising from the individual's own concerns, knowledge gaps, or new abilities, extending beyond advancement of an existing task list.
The effects of psychological state need further examination too. Recording an emotion value or placing it in context does not by itself establish influence on a choice. Actual cognitive inputs, grounds for judgment, and behavior need to be connected to identify its contribution.
Important branches can be supported by comparisons: change relevant state or information while keeping other conditions as similar as possible and observe whether judgments differ explainably. Repeated observations, input records, and version control help distinguish state effects from other changes.
Reasonable non-action matters as well. The individual may defer because of a responsibility, insufficient expected benefit, a capability gap, or unsuitable timing. Overall validation still requires actual meaningful autonomous activity; continual refusal or waiting cannot replace it.
5. How learning, sleep, and continuity are validated jointly
Learning validation needs to connect what came before, what was experienced, what changed internally, and what happened later.
An experience can produce a knowledge revision or a method candidate awaiting evaluation. Validation must examine substantive content, supporting evidence, applicability, and evaluation results rather than only a “learning complete” flag.
Later situations should test transfer. The individual needs to judge applicability itself and use relevant content. In similar but inapplicable situations, it also needs to adapt or abandon the original method.
Sleep validation observes how natural activity creates consolidation needs, what work actually completes, and how results enter life after waking. Unfinished consolidation needs to remain unfinished. A generated summary must be assessed from its actual content to determine what it accomplished.
Continuity runs through these processes. Can relevant experience be retrieved in a new context? Are responsibilities retained while awaiting other matters? Is new understanding used after sleep? Can psychological state, memory, and unfinished work continue after process interruption?
These situations can connect within one life history, but must not become a mandatory script for every cycle. An activity may create no new commitment or skill. Sleep, waiting, and interruption occur under actual conditions.
6. What different evidence can establish
Development requires several checking methods, each answering different questions.
| Evidence level | Main questions it can examine |
|---|---|
| Static review | Responsibilities, interfaces, provenance rules, and known gaps in design and code. |
| Module tests | Numerical computation, state transitions, transactions, duplicate handling, and specific failure conditions. |
| Controlled clock | Long-term state dynamics, time units, segmented computation, and recovery consistency. |
| Local live operation | Formal startup, actual models and durable storage, real capability calls, and result read-back. |
| Natural residence | Activity formation without direct prompting, changing concerns over time, natural sleep, and continuing life. |
| Live Linux operation | Startup, paths, permissions, storage, process control, and same-individual recovery in the deployed environment. |
A controlled clock can help check long-term mechanisms, but accelerated time needs explicit identification. It cannot create actual relationship, learning, or work experiences.
Controlled model responses, human-authored materials, and fault injection also need their conditions stated separately. Even with real execution paths and databases, prearranged semantic results support engineering conclusions only within their corresponding scope.
Evidence therefore needs versions, configurations, models, run times, initial state, and human intervention. If later runs change these conditions, the applicability of earlier conclusions needs reassessment.
Existing engineering progress and local passes still cannot replace joint M1 acceptance. Specific verified scope should be read item by item from the latest implementation status and its corresponding evidence.
7. How failures, counterexamples, and uncertainty enter conclusions
Validation needs to examine conditions in which a mechanism should not have an effect.
For example, does the individual assume closeness without real relationship evidence? Does external material claiming “you now have this ability” directly overwrite self-understanding? Does it act from old memory after permission is revoked? Does repeated arrival of the same event create satisfaction, commitments, or execution effects again?
These counterexamples help establish whether boundaries actually hold. They need to accompany positive cases: correctly refusing an unauthorized operation must coexist with performing necessary action under legitimate conditions.
Fault validation also needs important interruption points, including failed saving, execution before confirmation, interrupted background learning, and failed sleep consolidation or wake continuity. Recovery needs to check original state, outstanding obligations, and actual results.
Conclusions should distinguish not run, insufficient evidence, partial support, pass, and failure, with limitations. Rerunning after a fix can produce new evidence, while preserving the original failure and conditions.
If an expected change is absent, its cause also needs investigation. The mechanism may not have contributed, conditions may have been insufficient, or maintaining the prior state may have been reasonable. Experiences must not be invented, completion flags changed, or failures hidden to obtain an expected result.
8. How technical acceptance connects with public observation of Adam
Technical acceptance establishes capability scope, underlying records, and failure conditions. Adam's public records allow others to understand and discuss a specific individual's experiences over time.
Public records can show why something received attention, what choice the individual made, what actions it took, what resulted, and whether it changed later. Corresponding artifacts or appropriately prepared evidence can be supplied where needed, with human introduction, editing, and intervention identified.
Event time and publication time need separate records. Selected public records also cannot automatically represent the complete operating history or full acceptance scope.
Stage judgments need the same clarity: M1 examines the seven life capabilities and their interaction; M2 extends reliable work capabilities and associated safeguards; M3 adds Linux revalidation and longer observation. More runtime does not automatically yield a pass. The necessary evidence still needs to hold.
We aim to build an inspectable record of how the same individual generates mental activity from its state, acts, experiences and learns, and continues life with those changes.