AI research that makes capable models cheaper to run and easier to trust. https://ymirlabs.tech/log 2026-10-18T00:00:00Z Oviezibe Gift Jack © 2026 Nutin Technologies https://ymirlabs.tech/log#site-launch 2026-10-01T00:00:00Z

Ymir Labs has a home: what we’re researching, what Candid is meant to do, where we are, and REN’s results in full.

<p>This site sets out the question the lab works on, the outcomes Candid is designed for, and an honest account of our stage: early, pre-product and pre-revenue.</p><p>Every claim on it carries a label that says whether it has been measured, is in research, or is planned. We will keep this log as the dated record of what changes.</p><p><a href="https://ymirlabs.tech/how-we-work">How we work</a></p> https://ymirlabs.tech/log#ren-round-6 2026-10-01T00:00:00Z

The sixth training round of REN ran on the same free hardware as the first five.

<p>Round six was trained on Kaggle’s free tier (2× NVIDIA T4), like every round before it. Its runs and evaluations are on the public training log.</p><p>The figures quoted on this site are from round five.</p><p><a href="https://ren-training-log.vercel.app">REN training log</a></p> https://ymirlabs.tech/log#ren-round-5 2026-10-01T00:00:00Z

Evaluated on a held-out test set against the untouched base model, with 95% bootstrap confidence intervals.

<p>Against the untouched Qwen3.5-9B base model, evaluated in the same session, REN was clearly better on terminal tasks (42/42 against 36/42), quick judgements, calibration, structured-format answers, format compliance and tool-using agent turns. General knowledge showed no clear change, which means no loss.</p><p>Handling the interruption itself moved from 33/42 to 35/42. The confidence interval includes zero, so we count it as within noise: not a win yet.</p><p>The round ran for 740 optimiser steps over 20,441 examples and 9.51M tokens. The session cap cut it at step 646; it resumed from checkpoint with no skipped or NaN steps.</p><p>REN is a solo proof of concept. Candid is a separate architecture, to be built from scratch once the lab is funded.</p><p><a href="https://ymirlabs.tech/ren">Full results</a></p><p><a href="https://ren-training-log.vercel.app">REN training log</a></p> https://ymirlabs.tech/log#ren-rounds-1-4 2026-10-01T00:00:00Z

The first four rounds of REN, a solo proof of concept trained and evaluated on free GPUs.

<p>REN is a LoRA fine-tune of the open Qwen3.5-9B model, trained so it can be interrupted mid-answer and keep working sensibly. Its first four rounds ran on Kaggle’s free tier.</p><p>REN is a solo proof of concept by the founder. It is not Candid, and Candid is not built on it.</p><p>Every run and evaluation is logged publicly on the training log.</p><p><a href="https://ren-training-log.vercel.app">REN training log</a></p>