Tasks

Create from scratch a minimal engine that is easy to enhance using AI, featuring end-to-end typing, modular architecture, a minimalist frontend, and open extension cases.

Explore representing the entire task pool as a unified graph with AI folding into a human-comprehensible picture.

Explore how to monitor the health of dozens and hundreds of sites and quickly identify exceptions that require attention.

Explore the minimum requirements and the process for including an existing or new website into a unified management perimeter.

Determine what information about each website the central agent needs for discovery, routing, and batch operations.

Investigate ways to reduce the cost of technical changes that need to be deployed simultaneously across numerous independent websites.

Research a unified way to set and coordinate actions on individual sites and groups of sites through a common perimeter.

Investigate how to locate affected content and websites when new knowledge emerges or existing knowledge changes.

Figure out how to keep the central information contour up to date without regular full re-reading of all websites.

Investigate the central hub through which an agent can view the information space of multiple websites without constantly crawling them.

Understand what compact description of its own knowledge each website should be able to provide to operate within a large cluster.

Formulate and research a general approach to managing a growing cluster of websites as a unified informational and operational space.

Port the approach with reproducible verification scripts to the core: the agent must be able to use them as a definition of done and execute them independently after making changes.

Resolve incompatibilities between react-markdown for reading and @mdxeditor/editor for editing: a single Markdown content must be interpreted identically in the editor, client-side renderer, and SSR.

Add proper handling for the permanent deletion of public pages without falling back to a standard 404.

Add a universal addressable variables/artifacts mechanism to haih-agent/fi1osof.ru so that external AI clients can save intermediate data once and then reuse them by ID in subsequent operations.