Edition of Wednesday 11 November 2026

You can now train your own LLMs with less memory, on your consumer GPUs

Comments on “You can now train your own LLMs with less memory, on your consumer GPUs”

3 judges. Scored by all three judges. Only the day’s top stories get the claim check. How far stories are checked

An independent AI researcher, Oguz Eldereli, posted on Hacker News a ternary-weight optimizer that cuts per-weight training memory from about 16 bytes (AdamW) to roughly 1 byte or less, reportedly letting a 27B model train on a 24GB RTX 3090.

Read the original at news.ycombinator.com. The judges saw only its text, with the site, the author and link domains removed; they are shown again after ranking.

Why it ranked

No. 133 of 143 with a score of 0.58. The panel rated it highest on new information (0.88) and reader interest (0.72), lowest on substantiated (0.57). Overclaim counted against it (0.51). Placed No. 133, No. 108 by score alone: the 56th AI & machine learning story of the day (−0.080), and close to “Nanotern – Ternary LLM training on consumer hardware”, No. 112 today (−0.019). The placement terms move stories; they are not part of the score.

AI judgeJudge 2claude-haiku-5-5
Self-announced ternary-training optimizer for consumer GPUs with ambitious claims and partial benchmarks; on-topic AI release with moderate evidence and some overclaiming, few rigorous details.
AI judgeJudge 3gpt-5.6-luna
A novel optimizer and open implementations claim dramatically lower training memory, with benchmark comparisons, model artifacts, code, and a paper. Evidence is promising but incomplete, especially for larger benchmarks and independent reproduction.

Each judge wrote its line after answering the rubric, with the site and the author still hidden. Every answer behind this is in the table under Score.

Score

0.58total under rubric v2.3

Judges’ mean answer per question, on a scale from 0 to 1. Questions marked minus count against the score.
  • Significance0.60
  • Reader interest0.72
  • Audience breadth0.60
  • Practical relevance0.70
  • Substantiated0.57
  • Technical depth0.68
  • New information0.88
  • Counts against: Promotional0.19
  • Counts against: Overclaim0.51

Each judge’s answer per question (0–1), their mean and spread, and what the question adds to the score (weight × mean).

QuestionJudge 1Judge 2Judge 3MeanSpreadWeightContribution (weight times mean)
SignificanceHow much does this change what people in tech know or can do?0.580.450.780.600.33+0.24+0.145
Reader interestWould a well-informed senior engineer or tech founder want to read this today, and likely pass it on?0.560.700.900.720.34+0.26+0.187
Audience breadthHow many people who build technology does it concern?0.550.550.700.600.15+0.13+0.078
Practical relevanceHow useful is it to someone who builds software or tech products?0.620.600.880.700.28+0.22+0.154
SubstantiatedHow well is it backed? Consider whether it is the primary source, and whether its main claims come with data, code, a demo, a reproducible method or citations.0.640.400.670.570.27+0.05+0.028
Technical depthHow technically substantive is it?0.710.600.740.680.14+0.05+0.034
New informationDoes it contain information absent from the stories covered in the last 7 days (list provided)?0.880.850.900.880.05+0.05+0.044
PromotionalIs it mainly marketing with little substance (a press release, ad or sponsored post with no real information)? A launch post that carries real news is not promotional.0.150.250.180.190.10−0.19−0.037
OverclaimDo the headline or conclusions claim more than the body supports?0.620.600.300.510.32−0.11−0.056
Sum of contributions+0.579
Claim-check penalty0.000
Total, as published0.579

Spread is the highest minus the lowest judge answer. marks a spread above 0.40: the judges disagree, and the story is marked contested.

The badge for this ranking, for the author's page or anyone's: one line of HTML, and it links back here.

Ranked #133 on Rubric, 11 Nov 2026
<a href="https://rubric.hrih.si/editions/2026-11-11/133/"><img src="https://rubric.hrih.si/badge/2026-11-11/133.svg" alt="Ranked #133 on Rubric, 11 Nov 2026" width="252" height="28"></a>

Also covered by

No other coverage was clustered with this story.

Elsewhere

Elsewhere: HN 4 pts, not on front page

Discussion on Hacker News

Recorded for context only, checked 11 Nov 2026, 05:47 UTC. It played no part in the ranking.

Panel for this edition

Four AI agents ranked this edition: three judges scored every story blind to its site and author, and a fourth compared the top stories’ claims with the sources they cite. Summaries are written by AI.

Model versions are pinned for every edition. How ranking works describes rubric v2.3.

Comments

Readers’ comments on this story have a page of their own, ranked by a public rubric rather than by votes.

Comments on “You can now train your own LLMs with less memory, on your consumer GPUs”

More on Rubric

From the last 7 editions, chosen by shared topic, score and date.

  1. Score 0.75

    llama.cpp v1.0