Research-stage investigator

Every investigation
should make the
next one better.

Tend is an experimental investigator that builds working models of the systems it encounters. It tests what it thinks it knows, revises its understanding when evidence disagrees, and carries what it learns forward.

The ambition: a local investigator that keeps exploring and connects its learned models—little worlds of how things work—to become more capable through experience.

Inspect a real investigation How Tend works
Experience carried into the next investigation A winding path connects observations to retained experience, then branches toward a new question. Dotted alternatives remain open. ? OBSERVATIONSEXPERIENCEA BETTER NEXT QUESTION
Concept illustration: observations become experience and guide a better next question.

01 / How it works

Local models.
Actual observations.

Is Tend an LLM?

No. Tend’s current predictive core learns explicit models from structured actions and observed outcomes, without pretrained language-model weights. Predictions guide tests; actual observations support answers. Models and evidence remain inspectable and reusable.

Today, people still supply the interfaces and task boundaries.

How it works

Conceptual loop, not a live Tend run.

Act. Use the supplied interface to change or probe the system.

02 / One real result

One repair changed the next experiment.

In a macOS pipe investigation, Tend’s three-state model missed a distinction: an empty pipe could have notifications disabled. Actual observations contradicted it. Repair added a fourth state, which guided a useful later choice.

macOS pipe/kqueue · supplied actions and reset behavior

new acquisitions across four later service jobs

2repaired
model
3preceding
model

The repaired model skipped a failed drain-twice experiment and chose drain then enable directly.

An ordinary method using known system behavior tied it at two acquisitions. A complete-cost advantage remains unproven.

This is one selected native result, not a complete account of current research.

Read the result and its limits

03 / What’s next

Wider questions.
More useful experience.

We’re exploring whether learned models of systems—little worlds of how things work—can become a useful collection: alternatives Tend can test, connect and draw on during later investigations.

A longer-term ambition is for Tend to learn to understand and communicate about those worlds through experience.

Connected world models, learned language and general self-improvement are research ambitions, not established capabilities.

About Tend

A serious question?
Let’s investigate.

Have a concrete systems question with observable behavior and a way to tell whether the answer helped?

Built by Joe Kordish.

Discuss a research use case

Tend is a research project. A supported public release is still ahead.