Pierluigi Giuliana @Gaplus68

Agrigentino, milanista, zio, ingegnere, talvolta rompiscatole e (dicono) parco: il tutto mescolato, non agitato.. 'No one was born to be a servant or a slave' Agrigento Joined October 2010

Tweets

8K
Followers

836
Following

3K
Likes

5K

antirez @antirez

22 hours ago

The question is also *how do you ban* a blob of weights even within the US? It is impossible and the Internet is a proof of that at so many levels. People will go from a city to the other with hard drives, copy weights, share inference code. Hardware level protection is limited too.

Xiaoyin Qu @quxiaoyin

2 days ago

You cannot ban Chinese open source to make US win. The whole point of open source is it's OPEN. You might ban in the US, sure, but how are you going to ban it outside US? Instead the real question is, why can't US build the best open source model? Where is meta, gemini, @grok and

99 83 739 66K 58

27 11 232 24K 26

View Details

antirez @antirez

23 hours ago

Worrying thing with modern AI: GPT 5.5 explains me in hyper-GPU-jargon why something can't go faster than that. I finally understand its words, provide a hint on how to circumvent the problem, and it can get the win I expected. So GPT 5.5 is brilliant-jerk-ing every day more. Bad

29 15 397 30K 44

View Details

Pierluigi Giuliana @Gaplus68

18 hours ago

@HowardRoarkjr Se parla Prodi poi...

2 0 3 45 0

View Details

David Cramer @zeeg

20 hours ago

Does anyone have experience with any of the skill-specific eval frameworks? Would you recommend one? Needs to work with OpenRouter

7 2 28 7K 37

View Details

Pierluigi Giuliana @Gaplus68

4 days ago

@sono_selvatica @akashroy1k 6841

0 0 1 16 0

View Details

Pierluigi Giuliana @Gaplus68

a week ago

@Tecnobarbaro "era una normale procedura di collaudo"

0 0 0 10 0

View Details

Pierluigi Giuliana @Gaplus68

a week ago

@antirez It's a cult...

0 0 1 215 0

View Details

Computer Science @CompSciFact

a week ago

17 43 477 30K 46

View Details

antirez @antirez

a week ago

Generation speed in short contexts is much better, like 13.5 t/s. The generation speed number you see in the status bar is not really reliable, it is faster than that as you can see visually, but there is an upfront setup cost that must be reduced.

2 1 24 5K 1

View Details

antirez @antirez

2 weeks ago

I'm using the Unsloth 4-bit weights to integrate interence into DwarfStar. If the experiment works well I'll specialize the quants with an optimizer for the best setup. My target is however the 512GB M3 ultra and in distributed inference 4 128GB machines with 3-bit variants if quality holds. Inference quality must be the local inference mantra.

Unsloth AI @UnslothAI

2 weeks ago

GLM-5.2 can now be run locally!🔥 The 2-bit model retains ~82% accuracy after we shrunk it from 1.51TB to 238GB (-84% size). Run on a 256GB Mac or RAM/VRAM setups. GLM-5.2 is the strongest open model to date. Guide: unsloth.ai/docs/models/gl… GGUF: huggingface.co/unsloth/GLM-5.…

274 896 7K 1.8M 4K

24 21 421 40K 101

View Details

antirez @antirez

2 weeks ago

Redis users: would you like to have sorted sets using just 50% of the memory without any speed regression? I got a few ideas...

15 4 285 41K 28

View Details

antirez @antirez

3 weeks ago

What Europe should do right now: 1. Call all the European researchers working on AI and return them back with same salary (or they can stay but switch career). 2. Fill EU places having GPUs with money, and put those people there. 3. AI partnerships with China + India.

354 366 3K 209K 296

View Details

antirez @antirez

3 weeks ago

Took the good work of the communtiy of DwarfStar and consolidating the Strix Halo support. It looks very good. More QA in the next days and the final merge soon.

6 8 178 17K 18

View Details

antirez @antirez

4 weeks ago

Why I'm taking this DwarfStar thing so serious? It is from the times of Redis that this didn't happen. I believe strongly in local inference, as a safety net. But there is more: I enjoy doing this stuff. So at 50 you are still not wise enough to avoid doing new stuff. So lame.

33 43 807 39K 54

View Details

antirez @antirez

4 weeks ago

Redis Array!

Redis News Feed @redisfeed

4 weeks ago

[BLOG POST] Announcing Redis 8.8 redis.io/blog/announcin…

0 4 34 32K 13

6 11 204 27K 17

View Details

antirez @antirez

a month ago

We have the new Array type merged into Redis! github.com/redis/redis/co… If you have a use case or want to implement agents shared skills knowledge via ARGREP, this is the right moment to start :)

10 7 153 13K 22

View Details

Turing Post @TheTuringPost

2 months ago

An interesting attention mechanism from @AIatMeta: SP-KV (Self-Pruned Key-Value Attention) The model learns which tokens are likely to be useful for future attention and only keeps their key-value pairs in the persistent KV cache. For every token and attention head, a small utility predictor (a 2-layer MLP) computes a utility score, and older tokens are selectively pruned based on it. At the same time, attention becomes hybrid, because SP-KV still keeps a local sliding window fully available for short-range interactions This approach: - reduces KV cache size by about 3× to 10× - allows to compress longer contexts more - improves decoding speed and memory bandwidth

5 38 211 12K 146

View Details

@

2 months ago

Vi chiedo di fare il repost. Riguarda gli utenti che non riescono a postare, commentare, quelli che Twitter blocca e a cui compare il messaggio di spam, che sono spam. - Bisogna aggiungere un numero di telefono all'account - Oppure bisogna impostare il proprio account come satirico (io per ora ho solo impostato l'account satirico, senza telefono) Musk in questo momento sta ripulendo i bot, e in questa pulizia ci finiscono dentro anche gli utenti normali. Grazie a tutti 💜

2 79 90 2K 6

View Details

Ersin KOÇ @ersinkoc

2 months ago

WrongStack ve DeepSeek API kredisi ile kodlamaya devam. WrongStack sizi yolda bırakmaz, yalana dolana bulaşmaz, kodu piç etmez. :)

Ersin KOÇ @ersinkoc

2 months ago

ZAI, Kimi, MiniMax gibi coding/token plan kullanıyorsanız işte size yeni bir kod asistanı. Elbette API Key kullandığınız her provider ile çalışıyor. WrongStack kullandıkça aşık olacaksınız. :) Kurulum; npm i -g wrongstack@latest Key set etmek için; wrongstack auth Yolo Mod

10 5 29 16K 26

2 5 28 4K 24

View Details

antirez @antirez

2 months ago

Not bad. Committed.

antirez @antirez

2 months ago

If the correctness tests will pass, expect a sensibly faster DS4 inference speed for DGX Spark, and especially a lot flatter prefill as context increases. Soon in the repo if everything goes as expected.