Hacker News: GaggiX

New comment by GaggiX in "Gemini 4 Carbon"

GaggiX — Wed, 28 Oct 2026 18:57:09 +0000

They announced their new Index v5 yesterday, they frequently add new benchmarks and rebalance old ones.

New comment by GaggiX in "The new iPad mini"

GaggiX — Mon, 26 Oct 2026 14:23:32 +0000

With Apple ads there is always something uncanny with the hands, like the fingers get unnaturally longer behind the device.

New comment by GaggiX in "Dream-1"

GaggiX — Sat, 24 Oct 2026 19:42:46 +0000

This one as I read it is a bit different, the Physical Intelligence API requires finetuning for the task/robot while the Dream-1 model offers an API that can do zero-shot (with great limitations of course but still cool).

New comment by GaggiX in "Dream-1"

GaggiX — Sat, 24 Oct 2026 19:36:24 +0000

I feel like robotic tasks by LLM models and similar are very much underexplored, a few new frontier models are now able to do very simple tasks but hopefully there would be more competition in the future, this one was the first one that offers an API for this task specifically, maybe in the future if more models come up there will be a dedicated OpenRouter category, that would be cool.

New comment by GaggiX in "Dream-1"

GaggiX — Sat, 24 Oct 2026 19:15:03 +0000

I posted it after I discovered the model on https://robocurve.org/

I discovered the Robocurve benchmark after someone on X was saying that Kimi K3.1 Pro was performing particular bad on it.

Anyway I found the model particularly interesting because I think it's the first time I see an API for a random bimanual robot. Also HN removes "Introducing" so it's just the name of the model.

Dream-1

GaggiX — Sat, 24 Oct 2026 18:23:24 +0000

Article URL: https://dreamlabs.ai/research/dream-1/

Comments URL: https://news.ycombinator.com/item?id=50223785

Points: 180

# Comments: 77

New comment by GaggiX in "Is this the new cyber normal?"

GaggiX — Fri, 23 Oct 2026 12:05:53 +0000

LiveOverflow has exploited the PHP vulnerability, it’s even linked in the article.

New comment by GaggiX in "Mistral Large 4"

GaggiX — Wed, 21 Oct 2026 21:25:56 +0000

Have you set up some sort of benchmark to see if there is actually anything useful or they are complete trash on the task?

Sunsetting Decomp.dev

GaggiX — Thu, 15 Oct 2026 08:08:45 +0000

Article URL: https://decomp.dev/

Comments URL: https://news.ycombinator.com/item?id=50100731

Points: 9

# Comments: 1

New comment by GaggiX in "500B Tokens Later: Letting AI Agents Decompile a First-Person Shooter"

GaggiX — Sun, 11 Oct 2026 02:56:48 +0000

The vast majority of these tokens are cache input tokens so even at API price the cost would be much lower.

New comment by GaggiX in "Vibe coded browser ports of Halo, The Simpsons: Hit And Run, GTA work well"

GaggiX — Sat, 10 Oct 2026 19:33:20 +0000

The thing is they are actually good, that's why they are getting so popular even when you can emulate many of them already.

New comment by GaggiX in "Sub-1-Bit LLM Compression via Latent Factorization"

GaggiX — Thu, 08 Oct 2026 16:38:18 +0000

I recently found this 1.58-bit model for ASR and it's surprising good (and very fast), that being said it's not a LLM.

https://huggingface.co/moondream/parakeet-redux

New comment by GaggiX in "OpenDLSS: A Vulkan Reimplementation of Nvidia's DLSS 5 Neural Rendering Network"

GaggiX — Thu, 01 Oct 2026 09:16:04 +0000

Modders have added the ability to use the upscaler after DLSS 5, personally I don't know how sound this method is but the quality is pretty good, and allows to play games with DLSS 5 at 4k 60FPS with something that is not a 5090.

New comment by GaggiX in "OpenDLSS: A Vulkan Reimplementation of Nvidia's DLSS 5 Neural Rendering Network"

GaggiX — Thu, 01 Oct 2026 08:58:47 +0000

>they ship a model per game?

They don't. Only DLSS 1 was trained specifically per each game.

New comment by GaggiX in "When did Google get so weird?"

GaggiX — Sun, 27 Sep 2026 23:02:56 +0000

Reasoning models can do math on their own without external tools.

New comment by GaggiX in "Moondream Parakeet Redux and Parakeet Ultra"

GaggiX — Sun, 27 Sep 2026 08:54:13 +0000

I have tested the ternary model on English/Italian and I'm surprised by how good the model is, also it's very fast.

Moondream Parakeet Redux and Parakeet Ultra

GaggiX — Sun, 27 Sep 2026 08:52:00 +0000

Article URL: https://moondream.ai/blog/introducing-parakeet-redux-and-ultra

Comments URL: https://news.ycombinator.com/item?id=49864691

Points: 4

# Comments: 1

New comment by GaggiX in "94% on AIME with 1B Params"

GaggiX — Tue, 22 Sep 2026 15:23:23 +0000

Don't editorialize the title of the articles.

New comment by GaggiX in "Xiaomi Mimo 2.6 live post-training dashboard"

GaggiX — Wed, 16 Sep 2026 20:58:30 +0000

These are post-training reinforcement learning steps.

New comment by GaggiX in "Show HN: Pelican-bicycle alternatives"

GaggiX — Mon, 14 Sep 2026 16:13:41 +0000

I don't think a bunch of similar tasks can really saturate the "create a SVG of X", because the model should have a quite good spatial understanding of the world and how everything interacts.

For example Gemini 3.8 Flash seems very impressive at first glance but the results are not actually very coherent, this shows that its "world model" is not particularly great (compare to SOTA models).