HackerReposA FIELD GUIDE TO OPEN SOURCE
ProSponsor the shelf♧ 0
← Back to top all-time
HW-whistleblower
True-Story-of-Pangu
OPEN SOURCE2025–2025
HW-whistleblower / True-Story-of-Pangu

True-Story-of-Pangu

诺亚盘古大模型研发背后的真正的心酸与黑暗的故事。

2Distinct people
2Mentions
0Explicit recommenders
11.6KGitHub stars
REPOSITORY AT A GLANCE

What to know before exploring

What it is

诺亚盘古大模型研发背后的真正的心酸与黑暗的故事。

GitHub

Where it is used

No documented product use has been verified yet.

What people are saying

No recommendation phrase was found in the indexed comments.

No classified warning is attached to this page. Read the discussions before treating silence as approval.

Also in these discussions

No other repository is repeatedly linked in the same discussions.

This week on Hacker News

Nov 7–Nov 14 UTC · Direct repository links and separately labeled related project stories. Indexed through 11/14/2026, 5:04:57 PM UTC.

No direct repository links were found in the indexed HN discussions over the past seven days. Older discussions are available below.

Mentions over time

HN mentions by 3-hour interval2 dated mentions

Observed mentions: 2025-07-07 04:29:12 UTC to 2025-07-09 04:27:19 UTC.

012025-07-07 03:00:00 UTC to 2025-07-07 06:00:00 UTC (exclusive): 1 mentions; 0 submissions, 1 commentsJul 7, 03:002025-07-07 06:00:00 UTC to 2025-07-07 09:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-07 09:00:00 UTC to 2025-07-07 12:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-07 12:00:00 UTC to 2025-07-07 15:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-07 15:00:00 UTC to 2025-07-07 18:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-07 18:00:00 UTC to 2025-07-07 21:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-07 21:00:00 UTC to 2025-07-08 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-08 00:00:00 UTC to 2025-07-08 03:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-08 03:00:00 UTC to 2025-07-08 06:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 commentsJul 8, 03:002025-07-08 06:00:00 UTC to 2025-07-08 09:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-08 09:00:00 UTC to 2025-07-08 12:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-08 12:00:00 UTC to 2025-07-08 15:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-08 15:00:00 UTC to 2025-07-08 18:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-08 18:00:00 UTC to 2025-07-08 21:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-08 21:00:00 UTC to 2025-07-09 00:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-09 00:00:00 UTC to 2025-07-09 03:00:00 UTC (exclusive): 0 mentions; 0 submissions, 0 comments2025-07-09 03:00:00 UTC to 2025-07-09 06:00:00 UTC (exclusive): 1 mentions; 1 submissions, 0 commentsJul 9, 03:00

Counts come from the repository’s dated source evidence. Index cutoff: 2026-11-02 08:16:59 UTC.

View mention counts and date ranges
UTC ranges include their start and exclude their end. Select a period to filter the source evidence.
PeriodStart (UTC)End (exclusive, UTC)MentionsSubmissionsComments
Jul 7, 03:002025-07-07 03:00:00 UTC2025-07-07 06:00:00 UTC101
Jul 7, 06:002025-07-07 06:00:00 UTC2025-07-07 09:00:00 UTC000
Jul 7, 09:002025-07-07 09:00:00 UTC2025-07-07 12:00:00 UTC000
Jul 7, 12:002025-07-07 12:00:00 UTC2025-07-07 15:00:00 UTC000
Jul 7, 15:002025-07-07 15:00:00 UTC2025-07-07 18:00:00 UTC000
Jul 7, 18:002025-07-07 18:00:00 UTC2025-07-07 21:00:00 UTC000
Jul 7, 21:002025-07-07 21:00:00 UTC2025-07-08 00:00:00 UTC000
Jul 8, 00:002025-07-08 00:00:00 UTC2025-07-08 03:00:00 UTC000
Jul 8, 03:002025-07-08 03:00:00 UTC2025-07-08 06:00:00 UTC000
Jul 8, 06:002025-07-08 06:00:00 UTC2025-07-08 09:00:00 UTC000
Jul 8, 09:002025-07-08 09:00:00 UTC2025-07-08 12:00:00 UTC000
Jul 8, 12:002025-07-08 12:00:00 UTC2025-07-08 15:00:00 UTC000
Jul 8, 15:002025-07-08 15:00:00 UTC2025-07-08 18:00:00 UTC000
Jul 8, 18:002025-07-08 18:00:00 UTC2025-07-08 21:00:00 UTC000
Jul 8, 21:002025-07-08 21:00:00 UTC2025-07-09 00:00:00 UTC000
Jul 9, 00:002025-07-09 00:00:00 UTC2025-07-09 03:00:00 UTC000
Jul 9, 03:002025-07-09 03:00:00 UTC2025-07-09 06:00:00 UTC110

What Hacker News says

2 sources
Is the article a translation from Chinese? You have to have some deep knowledge on Chinese net slang and Huawei slang to correctly understand it. And all that unnecessary emotional expressions. All of it made the article hard to read. Here's takeaways I extracted: 1. The author claim to be "an employee of the Pangu Large Model Team and Huawei Noah's Ark Laboratory", a lower ranking "small worker". The first 4 bullet points supposed to prove that they have insider knowledge, which should authenticate the claims that followed. As of why Huawei named their teams in this oddly way is unexplained but do desire some psychiatric analysis. 2. "At the beginning, our (Huawei, editor's note) computing power was very limited..." (detail followed), "...At the same time, other domestic companies such as Alibaba (which published Qwen, editor's note) and
Read the rest
Zhipu were training on GPUs and had already figured out the right method. The gap between Pangu and its competitors was getting bigger and bigger" 3. "In this situation, Wang Yunhe ('the current director of Noah', editor's note) and his small model laboratory took action. They claimed that they inherited and transformed from the old 135B parameters, and through training a short few hundred B of data, the average improvement of various indicators was about ten points. In fact, this was their first masterpiece of applying the shell to the large model. Huawei's laymen led the experts, which made the leaders completely unaware of this nonsense. They only thought that there must be some algorithm innovation. After internal analysis, they actually used Qwen (which is published by Alibaba, editor's note) 1.5 110B for continued training.", "By adding layers, expanding the ffn dimension, and adding some mechanisms from the Pangu pi paper, they gathered about 135B parameters. In fact, the old 135B has 107 layers, while this model has only 82 layers, and the various configurations are also different. After training, the distribution of many parameters of the new 135B of unknown origin is almost exactly the same as that of Qwen 110B. Even the class name of the model code was Qwen at the time, and they were too lazy to even change the name. The subsequent model is the so-called 135B V2. This model was also provided to many downstreams at the time, even including external customers." And that's about it. Also, yeah, the article was indeed a translation from Chinese. The [original post] was written in Chinese, and then got translated it to English by github.com/moonlightelite. That's why it felt odd to read. [original post]: https://web.archive.org/web/20250706034203/https://github.co... After reading the article, I feel this is less of a whistle blowing, more of an attack against Wang Yunhe. That's why there's so much emotional expressions, to (maybe) appeal to Huawei and/or the future employer of this individual. But that's just my personal feelings/hint.
In Pangu's Sorrow: The Sorrow and Darkness of Huawei's Noah Pangu LLM R&D Process

MAKE YOURSELF AT HOME

A few shortcuts

/
Search the collection
?
Open this guide
Esc
Close dialogs and navigation

Your saved repositories and interests stay in this browser.