Kasai Labs · Research

Applied research and its current status.

Kasai Labs is an independent applied-research laboratory. We investigate real problems in learning, evaluation and trustworthy computation, and build the minimum intervention needed to learn from them. We plan to publish our findings as open-access preprints; the first is in preparation.

Every claim on this portal carries its real status. Work in preparation is labeled as such.

What this portal is

This is the institutional record: publications and experiments, written in English for academic peers, registries and indexes. It is one of three surfaces over the same work, and all three read the same status. The narrative of why we investigate something and where it ends up — in Spanish, for a general audience — is the laboratory log. The experiments you can actually run in your browser live in the laboratory app (prototype).

Research lines

Adaptive systems and psychometrics

How effective are vocational guidance instruments, and how do learning curves behave under adaptive sequencing?

Status: Experimental design

Multi-agent architectures

How does architectural drift emerge in LLM-programmed systems, and how can it be quantified?

Status: Active research

Neuro-symbolic AI and formal verification

Can axiomatic compression improve reliability in hybrid reasoning systems?

Status: Active research · doctoral line

Sovereign infrastructure and digital identity

What protocols enable selective disclosure and data sovereignty for individuals and institutions?

Status: Exploratory

How we publish

Our publication policy calls for preprints to be deposited with a DOI and released under CC BY 4.0. Affiliation is written as Kasai Labs. See Papers for the policy and preparation status, and Experiments for the current state of open work.