ljupco.bsky.social - Profile | ThreadSky | a Reddit-style client for Bluesky

Taking the Suwalki Gap is one of the higher probability options for Russia to test Europe's resolve. Salami tactics, always trying to maintain some crumb of justification. (Here, Kaliningrad isolation.) Rest of the Baltics is obvious next step after that.

submitted 11 days ago • 0 comments

I find the #MLSky feed is quite good these days, for both papers and news.

submitted 15 days ago • 3 comments

By learning you will teach; by teaching you will learn. Latin proverb

submitted 22 days ago • 0 comments

o3-mini is really good at writing internal documentation - feed it a codebase, get back a detailed explanation of how specific aspects of it work simonwillison.net/2025/Feb/5/o...

submitted 22 days ago • 6 comments

an MCP server so claude can use R1 for deep thinking github.com/martinbowlin...

submitted 34 days ago • 1 comment

the R1 effect

submitted 34 days ago • 5 comments

Not that you need another thread on Deepseek's R1, but I really enjoy these models, and it's great to see an open, MIT-licensed reasoner that's ~as good as OpenAI o1. A blog post: itcanthink.substack.com/p/deepseek-r... It's really very good at ARC-AGI for example:

submitted 36 days ago • 2 comments

Fewer journalists are training to be court reporters – that’s a problem for justice ~

submitted 42 days ago • 0 comments

A little comic about desire paths.

submitted 43 days ago • 86 comments

🏎️ Today I'm introducing a method to train static embedding models that run 100x to 400x faster on CPU than common embedding models, while retaining 85%+ of the quality! Including 2 models with training scripts, datasets, metrics, evals, ideation, all public. Details in 🧵

submitted 43 days ago • 1 comment

Word of the day: anthropectomy, meaning the opposite of anthropomorphism, when we mistakenly DENY that animals lack a human characteristic. Chimps, coyotes, boars, and others act in ways that imply they have a concept of death, as @susanamonso.com lays out in her new book. h/t @danfalk.bsky.social 🧪

submitted 54 days ago • 4 comments

Why, oh why can't Starmer see that economic growth depends on using the country's undoubted strengths in science, engineering and technology. Delegating the drive for growth to the bean counter in No.11 is a recipe for continued relative economic decline. www.ft.com/content/1834...

submitted 57 days ago • 0 comments

Whether you love or hate him, I do like how oddly transparent and frank some of 'A Journey' by Tony Blair is. Here's a segment in which he admits that he didn't really want to ban fox hunting, which is why the resulting ban was a bit rubbish and not well-enforced. www.amazon.co.uk/Journey-Tony...

submitted 59 days ago • 2 comments

the easiest tool I've found for connecting with mutuals from Twitter is Sky Follower Bridge setup the Chrome/Firefox/Edge extension: github.com/kawamataryo/... make sure you generate an app password to use only for Sky Follower Bridge: bsky.app/settings/app...

submitted 66 days ago • 2 comments

📢 Sky Follower Bridge version 2.3.0 has been released! 🚀 ✨New Feature: Migrate following from Instagram to Bluesky www.sky-follower-bridge.dev/get-started-... ⚠️ The matching accuracy is not as high as X. Please always visually confirm the matched users.

submitted 63 days ago • 3 comments

weird cultural thing that's happened to me is that after 15 years of living in Britain I now parse French people as incredibly, irritatingly rude, they're technically my people but christ, would it kill them to say please and thank you and learn to modulate their tone with strangers

submitted 63 days ago • 46 comments

I think this is a quite general thing: IME, a decent chunk of people from more or less every culture who spend time in the UK and absorb its politeness norms come to find their home culture ruder. (At any rate, I’ve heard similar from as widely different people as Greeks and South Africans.)

submitted 62 days ago • 0 comments

DeepSeek v3 LLM is out! According to benchmarks it is as good as GPT-4o/ Claude 3.5 Sonnet, but open source. Here's a summary:

submitted 62 days ago • 4 comments

And now, DeepSeek v3 www.deepseek.com 👏👏

submitted 62 days ago • 0 comments

$5M to train a sonnet-class model? 🤔 github.com/deepseek-ai/...

submitted 63 days ago • 1 comment

Do you know you can entirely swap the Adobe software stack with open source? May the #foss be with you. #floss #education os-sci.com/blog/our-blo...

submitted 63 days ago • 3 comments

Haha so true!

submitted 67 days ago • 0 comments

Smarter, Better, Faster, Longer: A Modern Bidirectional Encoder for Fast, Memory Efficient, and Long Context Finetuning and Inference Introduces ModernBERT, a bidirectional encoder advancing BERT-like models with 8K context length. 📝 arxiv.org/abs/2412.13663 👨🏽‍💻 github.com/AnswerDotAI/...

submitted 69 days ago • 0 comments

A Survey on Large Language Model-based Agents for Statistics and Data Science Presents a comprehensive survey examining LLM-based data agents' evolution, capabilities and challenges in making data analysis more accessible through natural language interaction. 📝 arxiv.org/abs/2412.14222

submitted 69 days ago • 0 comments

Wow 👀

submitted 69 days ago • 0 comments

Google is on a roll lately! Gemini 2.0 flash model closed the gap to Claude 3.5 Sonnet while being a small model and super fast. Gemini-exp-1206 similarly good. Notebook-lm improvements. And now great video generation with #Veo2? 🤯

submitted 72 days ago • 2 comments

Improving smaller LLMs by giving them more "time to think" similar to o1 from OpenAI. With longer thinking, Llama 1B outperforms Llama 8B and Llama 8B outperforms Llama 70B. It's great to see smaller models punching above their weight!

submitted 72 days ago • 1 comment

gemini-2.0-flash-thinking-exp-1219 released. Google's version of the OpenAI o1 model. 12 days of GoogleAI. 🤓

submitted 70 days ago • 2 comments

I am thrilled to announce the release of ModernBERT, the long-awaited BERT replacement! There might be a few LLM releases per week, but there is only one drop-in replacement that brings Pareto improvements over the 6 years old BERT while going at lightspeed

submitted 70 days ago • 3 comments

BERT is BACK! I joined a collaboration with AnswerAI and LightOn to bring you the next iteration of BERT. Introducing ModernBERT: 16x larger sequence length, better downstream performance (classification, retrieval), the fastest & most memory efficient encoder on the market. 🧵

submitted 70 days ago • 1 comment

BERT is so back 🔥 Answer AI and Lighton released ModernBERT: lightning-fast state-of-the-art BERT model with Apache 2.0 license 🥹 2x fast as debertav3 and 3x faster than nomic 💨 all models are here hf.co/collections/answerdotai/modernbert-67627ad707a4acbf33c41deb read more hf.co/blog/modernbert 📖

submitted 70 days ago • 3 comments

I'll get straight to the point. We trained 2 new models. Like BERT, but modern. ModernBERT. Not some hypey GenAI thing, but a proper workhorse model, for retrieval, classification, etc. Real practical stuff. It's much faster, more accurate, longer context, and more useful. 🧵

submitted 70 days ago • 19 comments