RELDEN Development Update — September 21, 2026

At the same time, RELDEN continued to take shape as a product — from work surrounding RELDEN-8B to its public identity and website.

What changed this week

A major area of progress was RELDEN's system for using additional computing resources.

The long-term goal is for RELDEN to handle as much work locally as practical. For tasks that require more computing power, however, RELDEN should also be able to temporarily use additional resources.

This week, we tested the complete lifecycle of such a process under real conditions: a temporary computing environment is created, recognized and monitored by RELDEN, used for its intended task, and then safely shut down and removed again.

These tests uncovered several issues that were not visible during local testing. Communication between individual components, handling of timing information, and validation of successful test runs all required further work.

Rather than bypassing safeguards to make the tests pass, the affected components were corrected and tested again.

A complete lifecycle test was subsequently completed successfully without violating the required safeguards. The temporary environment was created, monitored, stopped, and removed as intended, with no active resource left behind.

Why this matters

For users, most of this technical complexity should eventually remain invisible.

RELDEN is not intended to become simply another interface for accessing different AI models. The broader goal is to build a system capable of deciding where a task can be executed appropriately while keeping resource usage controlled.

That requires RELDEN to reliably understand when local computing power is sufficient, when additional resources are required, whether an external task is still active, and when those resources can safely be released again.

The last point is particularly important because external computing resources create ongoing costs. Resources continuing to run after a task has finished would therefore be both technically inefficient and unnecessarily expensive.

This week's testing represents another step toward making that process controlled and reliable.

RELDEN-8B is approaching its next training run

Preparation for RELDEN-8B also continued.

RELDEN-8B is an effort to adapt an existing language model more closely to the way RELDEN works. The objective is not to build a large foundation model entirely from scratch. Instead, the goal is to teach an already capable model to work more effectively with RELDEN's workflows, tools, and decision structures.

The next training run, internally referred to as RUN002, has been prepared.

We are deliberately applying a strict rule here: a training run should only begin when the infrastructure required for it has passed a current validation process.

An initial attempt was therefore automatically prevented because the previous validation was no longer within the required time window. No training resources were created and model training did not begin.

After a fresh successful lifecycle validation, the infrastructure was cleared for the next stage.

As of this update, RUN002 is prepared, but the verified results covered by this Development Journal do not yet confirm a successfully completed training run.

We deliberately distinguish between “ready to run” and “successfully trained.” Results will only be presented as progress once they have actually been measured.

RELDEN's public identity is becoming clearer

Work also continued on RELDEN's public presence.

The website is moving toward a clearer visual identity built around RELDEN's distinctive blue R symbol and a dark, minimal design language.

The globe used in the main presentation of the website reflects an important part of the project's direction: local control as the foundation, combined with the ability to access additional infrastructure when required.

This Development Journal is part of that public presence as well.

Rather than presenting only finished features, the goal is to document what has actually been completed, what is currently being developed, and what remains unfinished.

RELDEN is developed under LST — Laubenstein Technologies.

Current development status

Several important foundations have now worked successfully in real tests.

Temporary computing resources can be monitored through their lifecycle, communication between the relevant components has been validated, and automatic shutdown and cleanup have been successfully tested.

The safeguards surrounding future training runs are also behaving as intended — including preventing execution when required validation is no longer current.

RELDEN-8B RUN002 remains in the next stage of development. Its preparation has progressed and the required infrastructure validation has been completed, but a successfully trained and evaluated RUN002 model is not yet part of the verified results covered by this update.

RELDEN as a whole also remains under active development. The current website, infrastructure tests, and model work demonstrate the direction of the project, but they should not be interpreted as a finished product.

What's next

The immediate technical milestone is RUN002.

After training, the important question will not simply be whether the model can be trained successfully. We need to determine whether the adaptation actually improves RELDEN's intended capabilities without significantly damaging the broader capabilities of the underlying model.

The resulting candidate will therefore be evaluated against predefined tests before any conclusions are made about its usefulness.

In parallel, RELDEN's public presence will continue to be refined. The RELDEN identity, Development Journal, and existing visual direction will be brought together into a more consistent product experience.

Much of this week's work focused on things users should eventually never have to think about: resources starting reliably, safeguards doing their job, failures being detected, and unnecessary infrastructure disappearing when its work is finished.

Those invisible foundations are necessary before RELDEN can reliably take on larger and more autonomous tasks.

LST — Laubenstein Technologies

Building RELDEN — A More Capable You.