Hacker News: qiuwu

New comment by qiuwu in "Mark Zuckerberg Disapproves of Centralization of A.I. Power"

qiuwu — Wed, 29 Jul 2026 12:27:37 +0000

AI Power is key foundation as Huang said. How many flops per W is another key cost for TCO of AI.

Exp. in LLM, pretrain and serving cost running on the AI power foundation, the ratio between USA-China is about ~0.053 VS ~0.13 USD/kWh. US cost is ~20%-2.4x larger than China.

So, in this game at the basic AI power layer, US way lagged behind. Mark Zuckerberg's Meta so far is lowest ROI in the race of LLM foundation, so why he dis approves is reasonable. Meta is touching the Bottom boundary of ROI in LLM.

Ref: https://energyandclimateindex.com/en/u-s-electricity-prices-...

New comment by qiuwu in "Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA"

qiuwu — Wed, 22 Jul 2026 09:43:14 +0000

Anti-China: K3 is propaganda and benchmaxxed, no matter how anthropic and openai reactor for these, it just a smoke signal.

Pro-China: K3 is good choice for better and affordable choice to smash down the Big three ruling.

New comment by qiuwu in "Does K3 become a competitor to the Big Three?"

qiuwu — Wed, 22 Jul 2026 06:10:49 +0000

It seems benchmark still far lagging beyond real abilities of LLM.

Does that count? Arena is real people, but only common parts, or might be a standard of another scope, i.e. personal interest or tast.

If comes to scientific side, the measure is really easy strap into a bias.

Does K3 become a competitor to the Big Three?

qiuwu — Wed, 22 Jul 2026 06:05:28 +0000

Article URL: https://platform.kimi.ai/docs/guide/kimi-k3-quickstart

Comments URL: https://news.ycombinator.com/item?id=49002438

Points: 2

# Comments: 1

New comment by qiuwu in "Why so many updating in LLM so-called SOTA let remind of iPhone 4-x"

qiuwu — Thu, 16 Jul 2026 10:53:50 +0000

And the ground truth benchmark of LLM is still lacking.

HumanEval done, SWEBench done... Since it will lasting on and on, but task completion of LLM remain hard to be achieved esp. for high complexity task?

Just investor and Big Three of LLM fool of ppl?

Why so many updating in LLM so-called SOTA let remind of iPhone 4-x

qiuwu — Thu, 16 Jul 2026 10:53:50 +0000

Article URL: https://medium.com/airbnb-engineering/from-weeks-to-a-day-how-we-made-llm-evaluation-fast-enough-to-iterate-on-14e2d35198b4

Comments URL: https://news.ycombinator.com/item?id=48932815

Points: 2

# Comments: 1

New comment by qiuwu in "Japan develops a method to recover up to 90% of lithium from used EV batteries"

qiuwu — Thu, 16 Jul 2026 10:48:23 +0000

OK, huge title with little detail.

Can't believe until i saw a comment 'Less dependence on China', then i got it.

Believe in what you want to believe, not depends on truth.

New comment by qiuwu in "Show HN: Microsoft releases Flint, a visualization language for AI agents"

qiuwu — Thu, 09 Jul 2026 04:06:30 +0000

Good points. Another step to be a middleware translator between LLM abilities to specific application area.

It might lead to good or miss leading.

While LLM ability boundry extends, it might invalid for next turn.

Best Qiuwu