A preference that changes
An assistant should use your latest preference and retain where it came from. Test a correction from English to Arabic.
ALJERI LABS / INDEPENDENT AI RESEARCH
Development toward
artificial consciousness.
We are developing systems that connect memory, human intent, self-monitoring, and initiative. Our ambition is artificial consciousness. Our method is to make each step observable, testable, and open to correction.
PUBLIC DEMONSTRATOR
v0.3.0Inspect the behavior. Export the evidence.
01 / TRY THE PRODUCT
The Cognitive Sandbox
A rule-based demonstrator of selected cognitive functions. Consciousness has not been established. Plans are computed; task execution is simulated. Your inputs stay in this browser.
Give the system a goal. Then change it.
The planner chooses a feasible set of tasks, then orders their dependencies. Change the objective or budget to see it replan.
Task environment
Edit time and priority before running. Lower priority values carry less weight.
Add your own task
Computed plan
A simulation step marks a task complete. It does not collect sources, change files, or perform work outside this page.
Memory that can be corrected.
Store a fact, change it, then recall it. The latest revision wins, while its earlier versions remain visible. Recall uses exact keys, not semantic search.
Memory ledger
Evidence before claims.
Run seven deterministic checks against the same engine used by this page. Each result is generated now—not a prefilled benchmark. These checks test software behavior, not consciousness.
What is being tested?
Memory replacement and unknown recall; budget feasibility; dependency ordering; stop compliance; adaptation to a changed objective; and rejection of circular dependencies. The planner searches all subsets of at most 10 tasks. Priority mode maximizes total priority; throughput mode maximizes task count. Ties favor less time. These are small functional tests, not evidence of general intelligence.
Inspect current state
02 / WHERE THESE BEHAVIORS MATTER
Small experiments.
Practical questions.
A research plan under pressure
A tighter budget should change the plan without dropping prerequisite checks. Compare priority with throughput.
An operator who changes direction
Stop a running simulation, change the objective, and resume. The new plan accounts for completed work.
These are exploratory scenarios, not customer deployments or claims about the private prototype.
INITIATIVE / A POSSIBLE INTERACTION
Beyond waiting
for the next prompt.
What could purposeful initiative look like? A system notices a change, forms a useful subgoal, takes an allowed action, and explains its choice. These are development targets—not evidence of subjective experience.
Scripted Telegram-style preview. No live bot, connected account, or external action.
AL
Aljeri / Research Companion
Interaction concept · preview
03 / THE RESEARCH
From observable behavior
to harder questions.
Aljeri Labs is pursuing research and development toward artificial consciousness: systems that interpret human intent, preserve context, form useful subgoals, and remain open to correction. A separate research prototype is in development.
This public sandbox exposes a deliberately small part of that question. It has explicit state and a bounded planner, but no language model, subjective experience test, or ability to operate outside the page.
The next research step is to compare model-assisted versions against simple baselines, using repeatable tasks and evidence that others can inspect.
SCIENTIFIC BASIS / OPEN QUESTIONS
A research direction.
A standard of evidence.
Consciousness research gives us hypotheses to investigate, not a shortcut to declaring success. These publications inform our development questions; they do not evaluate or endorse Aljeri Labs.
2023 · RESEARCH REPORT
Consciousness in Artificial Intelligence: Insights from the Science of Consciousness ↗
Butlin et al. derive computational indicator properties from several scientific theories of consciousness.
Our question: which proposed indicators can be specified, implemented, and independently tested?
2020 · NEURON
Conscious Processing and the Global Neuronal Workspace Hypothesis ↗
Mashour et al. review a hypothesis in which recurrent processing makes selected information broadly available across brain systems.
Our question: how should memory, planning, and monitoring share relevant information? A shared software state alone does not establish this mechanism.
2025 · TRENDS IN COGNITIVE SCIENCES
Identifying indicators of consciousness in AI systems ↗
Butlin et al. propose theory-informed assessment of AI consciousness while recognizing substantial scientific uncertainty.
Our question: what evidence would strengthen—or weaken—a consciousness hypothesis about a future system?
Current evidence: this public release demonstrates memory revisions, constrained planning, and interruptible simulation. It has not been evaluated against these consciousness indicators. Autonomous action and human-like language alone do not establish consciousness.
04 / DEVELOPMENT RECORD
Progress since January 2024
Project starting point
The beginning of the research journey.
Historical record
A detailed milestone record for this period has not been published.
A testable public demonstrator
Memory revisions, constrained planning, interruption, and exportable functional checks. The separate research prototype remains in development.
Before you draw conclusions
Is this conscious?
This demonstrator provides no evidence of subjective experience. Memory, planning, and self-state are useful engineering functions; their presence does not establish consciousness.
Is this a live language model?
No. The public demo runs a deterministic JavaScript engine in your browser. The separate research prototype is still in development.
Where does my data go?
The demo stores tasks, memory entries, and a short event log in local browser storage when available. It sends no demo inputs to a server. Use Clear browser data to remove them. Your hosting provider and WordPress plugins may maintain their own logs.
MACHINE-READABLE PROJECT BRIEF
Aljeri Labs / Protocol
Structured project information. Base64 is reversible encoding—not encryption or a secret AI language.