Sagas

Bringing the abilities of a model together.

Sagas explores how memory, relationships, reasoning and checking can work together within one model, with each contributing to a coherent whole.

Kind
Research programme

From individual discoveries to a working model

Sagas began with a question: how can advances in different aspects of intelligence become abilities that work together?

Our research had established achievements in retaining information and checking results against what supports them.

We built Sagas to investigate how these achievements could contribute to a model's continuing work.

This made their connections central to the research. An ability had to retain its purpose while becoming part of a larger whole.

Preserving information through the whole process

The first integrated experiments demonstrated that Sagas could process real information, preserve the resulting internal states and recover them exactly.

The recorded results showed faithful recovery.

Exact recovery remained intact during the model's operation.

Information could pass through a working model and remain faithfully available afterwards.

This gave the research a dependable basis for examining how several abilities could develop together without losing the information on which they depended.

Working across different information

Further experiments tested that preservation across prose, source code, structured records, word lists and mixtures of content.

Every internal state checked for exact recovery was reconstructed without error across the recorded material and larger batches.

These results established consistency across the material and sizes. They also allowed us to study how the character of the information affected the model's behaviour while its ability to preserve that information remained intact.

Checking what remains supported

Sagas also brought checking into the model's ongoing work.

In a recorded experiment, a previously accepted result remained supported when it was reconsidered against later information.

A result could be accepted, retained and reconsidered during the model's work.

This is central to Sagas: keeping useful abilities connected to what supports them as the model develops.

Developing together

Sagas continues to investigate how its abilities can contribute more fully to one another.

The integrated experiments provide a foundation for studying richer relationships, more demanding reasoning and learning that preserves continuity. Sagas studies how progress in one part of the model becomes useful throughout the whole.

Preserving what matters. Checking what holds. Bringing separate abilities into a developing intelligence.

Cite this record

Forlais Research (2026). Sagas. Forlais Research Library, record FR-P-09.

@misc{fr-p-09,
  author = {{Forlais Research}},
  title = {{Sagas}},
  howpublished = {Forlais Research Library, record FR-P-09},
  year = {2026}
}