Hacker News: KaoruAoiShiho

New comment by KaoruAoiShiho in "ARC-AGI Leaderboard"

KaoruAoiShiho — Sat, 25 Jul 2026 12:36:56 +0000

Appears to be benchmaxxing

https://x.com/quietnning/status/2080786711861407883

New comment by KaoruAoiShiho in "Nvidia, CoreWeave, and Nebius: Inside the Circular Financing of the GPU Boom"

KaoruAoiShiho — Sat, 11 Jul 2026 18:59:22 +0000

You're probably just responding to the headline but this person is an AI bull and isn't claiming it's a big deal, she's going into it and explaining it.

New comment by KaoruAoiShiho in "GPT‑Live"

KaoruAoiShiho — Wed, 08 Jul 2026 17:32:02 +0000

Click through to the link, the answer is no it uses the latest gpt models now.

New comment by KaoruAoiShiho in "Most arguments are about ego, not ideas"

KaoruAoiShiho — Wed, 01 Jul 2026 15:39:25 +0000

I think the parent's point is that if you are genuinely open to losing, the arguments can be productive because you can learn something instead... So stopping arguments is just another way of closing yourself off.

New comment by KaoruAoiShiho in "GLM 5.2 beats Claude in our benchmarks"

KaoruAoiShiho — Sun, 28 Jun 2026 22:39:02 +0000

And before you know it, you invented some openrouter provider from first principles...

New comment by KaoruAoiShiho in "GLM 5.2 beats Claude in our benchmarks"

KaoruAoiShiho — Sun, 28 Jun 2026 22:37:24 +0000

Are you sure fireworks is unquant? It's not listing precision on openrouter like everyone else.

New comment by KaoruAoiShiho in "GLM-5.2: The Most Powerful Open Model yet and the Brutal Reality of Running It"

KaoruAoiShiho — Fri, 19 Jun 2026 02:29:05 +0000

Terrible zero value article, I am extremely surprised it is upvoted.

That being said Artificial Analysis just came out with a brand new benchmark where it scored between opus 4.8 and gpt-5.5 and well behind fable-5 so it's definitely frontier-ish https://x.com/ArtificialAnlys/status/2067744637155226101

New comment by KaoruAoiShiho in "GLM-5.2 is the new leading open weights model on Artificial Analysis"

KaoruAoiShiho — Wed, 17 Jun 2026 13:26:56 +0000

This is really held back by one bench (omniscience accuracy) where it's really very far behind otherwise i think it's got at least a couple of points higher.

New comment by KaoruAoiShiho in "Running local models is good now"

KaoruAoiShiho — Tue, 16 Jun 2026 19:10:27 +0000

Fable largely fixed the annoying chatterness so sucks that it's gone now.

New comment by KaoruAoiShiho in "Show HN: Number Gacha, a gacha game distilled to its essence"

KaoruAoiShiho — Tue, 19 May 2026 22:59:24 +0000

Then that's just a video game might as well as play a video game why limit yourself to still confined to the rules of chess.

New comment by KaoruAoiShiho in "Cursor Introduces Composer 2.5"

KaoruAoiShiho — Tue, 19 May 2026 12:24:14 +0000

Kimi 2.5 has the best long context. For raw coding benchmark scores you can just post train on top of it with more specialized data. 2.5 is kinda old, 2.6 is the current release which is exactly just that and catches up to the frontier in most aspects.

New comment by KaoruAoiShiho in "Show HN: Number Gacha, a gacha game distilled to its essence"

KaoruAoiShiho — Tue, 19 May 2026 02:49:45 +0000

Do people like "gacha"? I thought people played games for the game experience, story, etc, and the gacha is just the monetization mechanic. It's like making a big deal out of paying $20 bucks a month, or buying loads of DLCs for example.

New comment by KaoruAoiShiho in "China blocks Meta's acquisition of AI startup Manus"

KaoruAoiShiho — Thu, 30 Apr 2026 12:48:10 +0000

No I think the best agent with hundreds of millions in ARR should be worth more than the 15th best model company with tiny revenue. ur the joke.

New comment by KaoruAoiShiho in "China blocks Meta's acquisition of AI startup Manus"

KaoruAoiShiho — Mon, 27 Apr 2026 19:29:07 +0000

Manus is saved, 2 billion is such an undervaluation considering much worse companies like minimax is valued at 30 billion.

New comment by KaoruAoiShiho in "DeepSeek v4"

KaoruAoiShiho — Fri, 24 Apr 2026 03:57:24 +0000

SOTA MRCR (or would've been a few hours earlier... beaten by 5.5), I've long thought of this as the most important non-agentic benchmark, so this is especially impressive. Beats Opus 4.7 here

New comment by KaoruAoiShiho in "Kimi K2.6: Advancing open-source coding"

KaoruAoiShiho — Mon, 20 Apr 2026 16:45:19 +0000

Huh, that's not a thing?

New comment by KaoruAoiShiho in "Claude Opus 4.7"

KaoruAoiShiho — Thu, 16 Apr 2026 15:57:46 +0000

Might be sticking with 4.6 it's only been 20 minutes of using 4.7 and there are annoyances I didn't face with 4.6 what the heck. Huge downgrade on MRCR too....

256K:

- Opus 4.6: 91.9% - Opus 4.7: 59.2%

1M:

- Opus 4.6: 78.3% - Opus 4.7: 32.2%

New comment by KaoruAoiShiho in "Pro Max 5x quota exhausted in 1.5 hours despite moderate usage"

KaoruAoiShiho — Sun, 12 Apr 2026 20:02:35 +0000

Talking nonsense.

New comment by KaoruAoiShiho in "Pro Max 5x quota exhausted in 1.5 hours despite moderate usage"

KaoruAoiShiho — Sun, 12 Apr 2026 17:17:48 +0000

Well they're not public yet so you'll have to put up with rumors. But the numbers are available for companies like DeepSeek say they have an 80% profit margin, so it stands to reason OAI etc would do similar numbers considering they charge much more.

New comment by KaoruAoiShiho in "Pro Max 5x quota exhausted in 1.5 hours despite moderate usage"

KaoruAoiShiho — Sun, 12 Apr 2026 14:14:40 +0000

After googling https://www.reddit.com/r/singularity/comments/1psesym/openai...