R&D · TRACK 05 / 05 · Experimental

Better from use,on your machine.

Off by default. Answers you like or correct can train a small LoRA adapter for that model, entirely on this machine, and only from turns you marked.

Read the notes View the code

What it does.
One system.

CONSENT · LIVE

Consent first

Nothing is used unless you marked that turn: a thumbs-up to keep an answer, or a correction to teach a better one.

OPT-IN

Local training

A LoRA adapter is trained on the original weights of that model, on your hardware. Nothing you wrote leaves the machine.

LORA

Checked before offered

The result is compared with the original model. If it got worse, you are not offered it.

CHECKED

Your choice

A new GGUF appears with a Use it button. Switch to it, or ignore it and keep the original.

USE IT

How it works

How it works.

Step 1

Mark

Thumbs-up good answers, correct the wrong ones.

Step 2

Collect

Marked turns become a filtered training set.

Step 3

Train

A LoRA adapter is trained locally on the original weights.

Step 4

Offer

If it passes the check, it appears with a Use it button.

Spec sheet.

01 / 05

Default

Default.

Off; enabled only when you ask

Default
Off; enabled only when you ask
Method
LoRA adapter on the model's original weights
Data
Only turns you marked
Where
Entirely on this machine
Status
Experimental, needs a separate training stack

In short.
On-device tuning.

Builds a dataset only from feedback you give

Checks the tuned model did not get worse before offering it

Exports a new GGUF you can switch to, or ignore

FROM IDEA TO RELEASE

Builders, not managers.

Say hello, fill in the blanks

Hi, I’m

and I want to talk about

.

Reach me at

Opens your email app, addressed to support@imutechz.in. Nothing is sent from this page.

or just email support@imutechz.in