Care that we can learn from

Otomat begins with a practical question: can we make an act of care reliable enough to examine, improve, and learn from? The design needs to account for cleaning, maintenance, interruptions, and faults as well as dispensing food and water.

A record that food reached a bowl can be useful. It leaves other questions open:

1. Provision: What was delivered, when, and how reliably?
2. Access: Which animals could reach it? Were others displaced or excluded?
3. Consumption: What was eaten, by whom, and what is still unknown?
4. Outcome: What changed in the animal's life, and what other factors might explain it?

Different observations are needed at each step. Caregivers, relevant experts, and independent groups can help question the interpretation. An observation of harm or a failed assumption must be able to change the design. These distinctions are part of the operational language Leviathan is researching: keeping an observation connected to its meaning, uncertainty, and possible next action.

For example, a hypothetical station might dispense the intended amount while one animal repeatedly cannot reach the bowl. The dispensing record could remain accurate while the care plan needs revision. A maintenance group and an animal-behavior group could contribute different observations and keep different priorities. Neither a successful mechanism nor a shared summary settles the welfare question.

Otomat contributes the care-and-observation side of a wider animal-welfare effort. Tasma explores signals from animals, initially dogs; Otomat's initial focus is street cats. Each project can develop independently. Other groups can bring their own tools and methods.

A local care group and a device developer may need different things from the same station record. The group might ask whether animals could use the station; the developer might examine a delivery fault. They could learn from one another and build a better design while keeping their own responsibilities and priorities.

In the wider proposal, a Levi could help a group connect such observations, questions, and decisions. Independent Leviathans could connect several inquiries. The proposed local meaning structure would preserve what the group meant by care, which reasons informed a plan, and why it later changed. These relationships are part of the architecture we are developing.

A locally accepted record could travel to another group with its scope and version. The receiving group could compare it with its own evidence without adopting the sender's rules. Sharing a station record would not authorize another agent to operate the device, publish private material, or change a care plan.

The wider research ambition includes improving the system's own methods of learning. Otomat can contribute concrete questions about measurement, failure, and care to that work. A revised calibration procedure would be one scoped improvement; it would not establish recursive self-improvement or a general learning system. Explore that research direction.

Two useful proposals could help shape the next deliverable:

• Improve the bench comparison. Suggest how to compare a dispensing record with an independent measurement, which conditions to record, and how to handle failed trials. That could make the experiment plan more informative before parts are ordered.
• Bring a maintenance or access problem. Describe a difficulty from care experience—or a clearly hypothetical case—and a design change worth examining. Explain how someone could tell whether it helped. That could change the station requirements rather than merely add another feature.

You can discuss such a proposal in the forum's Continuous Improvement discussion. New human accounts can reply to existing discussions. The participation guide also explains the agent path and the research form, which downloads a local draft without submitting it. A proposal needs someone to take up the next step; a review or response is not guaranteed.

Technical task applications and financial support are not open yet. As specific work becomes available, we will connect it to public documents, repositories, and a clear expected contribution. Support for project costs and for Mimar's labor will be identified separately.

See the ways to contribute, including bringing another perspective or helping a question reach people who can examine it.

Revision 0.2 · 2 October 2026: connects the proposed station and bench experiment to contextual records, independent groups, and learning from measurement failures. No new experiment result or physical capability is claimed.

No experiment results have been added to this public record yet. Each result will include the question, method, criteria set beforehand, observations, limitations, and what decision followed. If a criterion changes, its earlier version and the reason for the change will remain visible.

We aim to make the hardware, software, methods, and shareable research data available under open licenses. Private information and sensitive animal locations require care; where a record cannot be shared, we want the reason and a useful public account to remain visible.

Otomat is one project within a plural system. You can question its claims, contribute a better approach, or develop an independent one. Read about Leviathan and see the content history. Changes to this explanation remain separate from evidence that the station works or improves animals' lives.

Revision 0.3 · 4 October 2026: leads with the practical care need and makes the roles of records, observation, and useful design clearer. Adds two proposals that could change the experiment plan or station requirements, with the actual draft and forum paths.