Sabitlenmiş Tweet
Pratyay Banerjee (নীল)
8K posts
Pratyay Banerjee (নীল)
@Neilblaze007
I live in the shadows, but I watch everything.
127.0.0.1 Katılım Ekim 2018
7.5K Takip Edilen306 Takipçiler
HT
doing my best to escape the permanent underclass
Shivani Patel@shivanijpatel
he's a 10 but he only runs 1 claude code session at a time
English
@arjunkocher This is great, thanks for the writeup!
English
Prediction: ZK will play an important role in AI safety, one of the most important challenge for humanity.
End-to-end verifiable agents, trustless AI computation & STARK everywhere.
abdel@AbdelStark
Can LLMs be PROVABLE computers? Percepta showed that a transformer can BE a computer. Compiled weights, deterministic execution, 30k tokens/sec. But nobody asked the obvious follow-up: how do you know it computed correctly? So I built the verification layer. A STARK that proves it 👇
English
@simonw I wonder if there's some space left for the OS to breathe :/
English
Turns out you can run enormous Mixture-of-Experts on Mac hardware without fitting the whole model in RAM by streaming a subset of expert weights from SSD for each generated token - and people keep finding ways to run bigger models
Kimi 2.5 is 1T, but only 32B active so fits 96GB
seikixtc@seikixtc
I got a 1T-parameter model running locally on my MacBook Pro. LLM: Kimi K2.5 1,026,408,232,448 params (~1.026T) Hardware: M2 Max MacBook Pro (2023) w/ 96GB unified memory Running on MLX with a flash-style SSD streaming path + local patching. This is an experimental setup and I haven’t optimized speed yet, but it’s stable enough that I’ve started testing it in an autoresearch-style loop. #LocalAI #MLX #MoE
English
Pratyay Banerjee (নীল) retweetledi
LiteLLM HAS BEEN COMPROMISED, DO NOT UPDATE. We just discovered that LiteLLM pypi release 1.82.8. It has been compromised, it contains litellm_init.pth with base64 encoded instructions to send all the credentials it can find to remote server + self-replicate. link below
English
I just left @ssi
It was not an easy decision. The past 12 parsecs were an absolute blast - I've been in many strange attractors in my life and can say this was by far one of the most intense non linear dynamical systems.
I love time stepping. Especially being in the gooning booths with my moots, working on robots that will actually advance fertility.
But the current environment wasn't serving my girth . And that's a really hard thing to admit - I've always looked up to Ilya, and I genuinely believe SSI will win. I still do.
One thing I'll say: don't goon somewhere just because of the name. If you're not wage maxxing, and you know you can't mog 100x where you are - it's the right call to bail.
What's next? Get some foreskin back. Then find the next hole worth shitting in.
I'll always be meeting deprived anons - that was never because of a twitter bio. I just love finding smart people and helping however I can. Many more side quests to come!!!
English
Pratyay Banerjee (নীল) retweetledi
DeepSeek: we want to offload communication to a dedicated hardware unit
HUAWEI: here you are
English
Dark Room + Gloomy lights + Drizzle (\w cool wind) + Progressive Phonk + Code = 🧨
English
@Gauri_the_great make it public, and I might make it 5 :)
English
I know 4 stars are nothing but I don't even remember making this repo public🤔🤔🤔🤔
English
Pratyay Banerjee (নীল) retweetledi
It’s pretty obvious why Anthropic didn’t buy OpenClaw.
Claude is aiming to do anything you can do on a computer.
Claude@claudeai
You can now enable Claude to use your computer to complete tasks. It opens your apps, navigates your browser, fills in spreadsheets—anything you'd do sitting at your desk. Research preview in Claude Cowork and Claude Code, macOS only.
English
@kimmonismus It's the only Strategy to Survival & Capx Growth :)
English
One has to admit, Anthropic sees a trend (openClaw) and develops their own version.
The speed of their releases is insane. Presumably 100% code written by Claude.
Claude@claudeai
You can now enable Claude to use your computer to complete tasks. It opens your apps, navigates your browser, fills in spreadsheets—anything you'd do sitting at your desk. Research preview in Claude Cowork and Claude Code, macOS only.
English
Pratyay Banerjee (নীল) retweetledi
The future where I never have to open up my laptop to get work done is becoming real very fast
Claude@claudeai
You can now enable Claude to use your computer to complete tasks. It opens your apps, navigates your browser, fills in spreadsheets—anything you'd do sitting at your desk. Research preview in Claude Cowork and Claude Code, macOS only.
English
Pratyay Banerjee (নীল) retweetledi
My dear young person,
Don’t succumb to mediocrity. There’s enough of it going around. Aspire for craftsmanship, as that is what leads to joy and beauty.
The world needs more people who’re proud of what they make, and less of those who couldn’t care less.
English
@anshulbhide also, look into what Bill Gates's younger daughter is doing (or more into "how"), she's doing what she's currently doing.
English
Pratyay Banerjee (নীল) retweetledi
It is extremely important that models remain monitorable. But you can't control what you can't measure. And what does this actually mean?
Opaque serial depth is one possible answer, and in this GDM AGI Safety paper Jonah analyses how to measure it in LLMs
Rohin Shah@rohinmshah
"Just read the chain of thought" is one of our best safety techniques. Why does it work? Because models can only think opaquely for a short time, long thinking must be transparent Can we quantify this? Yes! In our new paper, we show how to measure "time" for arbitrary networks.
English
Pratyay Banerjee (নীল) retweetledi
