Site map
This site documents Serpentis, a research system that a human and an AI agent build and study together. It does not try to persuade or sell: it describes what has been done, what has been found, and how to check it.
The map serves two readers. A person can see the structure of the site at a glance. If you read the site with the help of an AI agent, give it a link to this page: it makes the site easier to find one's way around. The same map as plain text is at /llms.txt.
How to read what the site says
- On the History page, every finding states how sure we are: measured, observed (and in how many cases), a working hypothesis, or our decision. Dates there are given to the month and mark the first written record.
- The publications can be checked: each has a DOI, and the list is on the Publications page. The internal document the History was built from is open (see "Outside the menu" below).
- The system's code is closed. Where there is no link to a publication, what you read is our own account.
Pages
| Page | What it holds (the page's own description) | Text length | In Russian |
|---|---|---|---|
| Serpentis AI — a research and engineering system | Serpentis is an independent human–model research system for accumulating method, memory, and results. | ≈ 2k characters | /ru/ |
| Human–Agent Dyad: A Research and Engineering System | A high-level description of the Serpentis human–agent dyad and the research system built around it. | ≈ 24k characters | /ru/article |
| Aspects: what makes the system distinctive | The aspects that define Serpentis: dyadic work, memory, methodology, and layered verification. | ≈ 5k characters | /ru/aspects |
| Blog | Research notes on the human–model dyad, memory, methodology, and the open systems around them. | ≈ 2k characters | /ru/blog |
| The dyad: the unit that produces the result — not the human or the model alone | A Serpentis research note on the human–model dyad as the unit that produces the result. | ≈ 4k characters | /ru/blog_dyad |
| Memory that discards nothing: storage, the key, and the order of layers | A Serpentis research note on memory, storage, keys, and the order of layers. | ≈ 5k characters | /ru/blog_memory |
| Methodology is not a by-product — it is the product | A Serpentis research note on methodology as the product of a research system. | ≈ 4k characters | /ru/blog_methodology_as_product |
| Silicon psychology: why study the behavior of the “model + human” system | A Serpentis research note on studying the behavior of a human–model system. | ≈ 5k characters | /ru/blog_silicon_psychology |
| A tool that does what isn't its job: why a skill is an open system | A Serpentis research note on skills as open systems and tools that must meet their input. | ≈ 6k characters | /ru/blog_tools_as_open_systems |
| When it's all to spec, and it doesn't work: topological and functional integrity in practice | A Serpentis research note on topological and functional integrity in practice. | ≈ 8k characters | /ru/blog_topology_functional_integrity |
| Contact | Contact form and email address ([email protected]). | ≈ 1k characters | /ru/contact |
| Donate | Ways to donate to Serpentis and help sustain its research time, infrastructure, and public record. | ≈ 1k characters | /ru/donate |
| Product | Independent assessments of a Serpentis product currently in alpha. | ≈ 6k characters | /ru/product |
| Publications | Publications and archival records from the Serpentis independent research system. | ≈ 3k characters | /ru/publications |
| History: how the system grew | How the Serpentis research system grew: lines of research, what we found, and what the system does today. | ≈ 35k characters | /ru/history |
| Open source | The tools we built for our own work with AI agents. Each one solves a problem we hit ourselves. | ≈ 2k characters | /ru/github |
Outside the menu
- /ru/history_internal — System history, working version: the internal document the History was built from, without lines about people, plans and product ideas. Russian only; not indexed by search engines. ≈ 248k characters
- /llms.txt — This map as plain text, for programs.
The map is regenerated at every build of the site from the pages themselves: each description is the page's own, and each length is a count of its text. The site version is shown at the bottom of the page.