na8ve research
Research
na8ve is built on one idea: a child should have a learning model of their own, holding only what has been observed about them, readable and overrulable by the adult teaching them. We publish what that model does, run by run, including the runs where it was us who got it wrong.
Latest
Cognitive Architecture13 min read
If I Only Had A Brain: Cortex, an AI-native streaming datastore for memory, analytics, and auditability
Ask Cortex something, retract the fact behind it, and watch the answer change in the same conversation. A walk through the console built so a superintendent or a compliance officer does not have to take our word for what the cognitive layer is doing, including an unedited session recording of the console running: a streaming datastore built on neural networks and machine learning to think with what it holds instead of just storing it, one memory that serves as governance layer and analytics layer at once because both read the same reactive store, and a private institutional model that comes to represent the organization it serves, because its only source of truth about that organization is what the organization itself put there.
Read the article
How we publish
Predictions written first
Before every research run we write down what we expect the model to do and what result would prove us wrong. Our predictions are grounded in the model's prior runs, and established research.
Failures published in full
When a run goes badly we publish the run that went badly, with the defect record beside it. A program whose first output is a success story has not yet shown it can produce anything else.
A teacher stays in the loop
Nothing the model concludes grades a child, issues a credential, or gates what a student may do. The adult can read the reasoning and overrule it.
All research
Cognitive Architecture
If I Only Had A Brain: Cortex, an AI-native streaming datastore for memory, analytics, and auditability
The last piece described a cognitive layer in the abstract: an isolated memory per student and how it becomes an insights dashboard. This piece gives that layer a name, Cortex, and follows it from a single student's memory to an institution's: a streaming datastore built to think with the information it holds, and doubling as a governance and an analytics layer with its own memory and private model that comes to represent the institution.
David Laurenvil
Education
The Cognitive Layer Behind na8ve Models: Insights, Pedagogical Strategies, and Adaptive Learning
A single warm-up lesson is enough for na8ve to start deciding what a student needs next. What the model remembers about a student across courses, where that memory actually runs, what it decided when we put three students through the same skill on the same day, and why Stanford's evidence review says none of it works without a human still involved in the learning.
David Laurenvil
Education
Designing an Equitable Model for AI Education
A research-backed look at every major feature of na8ve, and what Stanford's 2026 evidence review says about why each one is necessary.
David Laurenvil
Education
What if every student had their own learning model, instead of a general AI tutor?
Better yet, what if parents and teachers had full authority and control of their child's individual AI, so they can see how and why their students learn and achieve?
David Laurenvil
The next one lands when the next research run closes, whichever way it goes.
See it on your own student
Everything in these articles runs in the same studio a parent or tutor uses to author courses and teach one student. You can open the model it builds, read how it reached a conclusion, and pin a different teaching approach when you disagree.
How na8ve works