Lukas

yapit.md turns urls and pdfs (even research papers) into clean, listenable markdown. @kepano's defuddle handles websites, a vision LLM handles PDFs and e.g. gives math spoken alt text. Free TTS in your browser, open source, self-hostable - and I didn't make UI/UX an afterthought.

@rfleury @cmuratori @dgtlgrove @croloris why not?

@lukasl_dev @cmuratori @dgtlgrove @croloris No thank you

Moving the @dgtlgrove community off Discord

Growing Graphs demo is finally out! 🕸️✨ 🔗 znah.net/graphs/ Videos from a few months ago finally meet a finished implementation, thanks Gemini for doing the boring parts. Inspired by Paul Cousin's Graph-Rewriting Automata: like a Game of Life, but cells can split if they want to #GenerativeArt #WASM #SwissGL

@rfleury @cmuratori @dgtlgrove we need awebo, @croloris

@cmuratori @dgtlgrove I have no idea yet :( part of me just wants to say screw it and retvrn to IRC and phpBB (or similar)

Post your Obsidian graph this is 5203 notes taken over 5 years

@adamwathan github.com/lukasl-dev/rime

What's the most useful MCP server you have installed?

cloud apps have locked generations of people into self-limiting beliefs about what a computer can do

@RhysSullivan @theo forgefed.org

@theo i want to see a federated version like bluesky where it's a network, matches the decentralized theme of git keeping an eye on tangled.org as it seemed cool but i think network effects are just too strong

It’s probably time for something new to replace GitHub

@julianhyde @MaineFrameworks There's a Google research paper discussing an alternative pipe-based syntax for SQL: research.google/pubs/sql-has-p…

@MaineFrameworks Both SQL and list comprehensions got it wrong. SQL should have put FROM first, because that’s the origin of the data (and therefore the types). List comprehensions seem “elegant” only because they are based on math notation en.m.wikipedia.org/wiki/List_comp…

"Left to Right Programming -- Programs Should Be Valid As They Are Typed"

Will you still need me, will you still feed me, when I'm 94? Happy Birthday Roger.

@VictorTaelin Isn't Deepmind interested in theorem proving given their recent Alpha* releases?

why does everyone in ML seems traumatized of formal methods? is there any AI lab, at all, passionate about it? you understand that, even if you don't think it is directly needed for AI, it is the one and only way to check math proofs in a computer - right?

@iamgiyu_ I'm using Tailscale for that

I need to build an SSH tunnel between my two laptops they both don't have any publicly addressable IPs and both are at different places, how can I do that ?

@charliermarsh ty

Today, we’re announcing the preview release of ty, an extremely fast type checker and language server for Python, written in Rust. In early testing, it's 10x, 50x, even 100x faster than existing type checkers. (We've seen >600x speed-ups over Mypy in some real-world projects.)

@charliermarsh amazing

We’re building a new static type checker for Python, from scratch, in Rust. From a technical perspective, it’s probably our most ambitious project yet. We’re about 800 PRs deep!

@VictorTaelin Generally speaking, I have the feeling that O1 tends to skip subproblems and marks them as "trivial". As soon as you ask more in-depth questions, it starts to fail. Maybe something like a type enumerator could help :)

Hey, the thing that makes me hate o1 doesn't happen in r1! Look this example, it is clear as sky. I wrote a prompt to teach the Interaction Calculus to the AIs, and I then asked them to reduce a simple term that requires grasping the core insight behind it. I asked Sonnet, o1-pro and r1. And... all got it right! It was a very simple exercise, but, still cool. Now, the actual answers? - Sonnet: `λt(t (λx0(x0)) (λx1(x1)))` - r1: `λt(t (λx(x)) (λx(x)))` - o1: `λt ( t (λx0. x0) (λx1. x1) ).` Can you see it? The problem is: o1, while technically correct, is the only that got the syntax wrong. This is not a logical error, it just said "fuck you" to my notation, invented its own thing and used that in the answer. And it is terrible. Perhaps I've been misjudging o1 by its inability to stick to the damn notation, but... why it does that?

🚀 DeepSeek-R1 is here! ⚡ Performance on par with OpenAI-o1 📖 Fully open-source model & technical report 🏆 MIT licensed: Distill & commercialize freely! 🌐 Website & API are live now! Try DeepThink at chat.deepseek.com today! 🐋 1/n

BREAKING NEWS The Royal Swedish Academy of Sciences has decided to award the 2024 #NobelPrize in Physics to John J. Hopfield and Geoffrey E. Hinton “for foundational discoveries and inventions that enable machine learning with artificial neural networks.”

My prediction for how coming AGI systems will look like: A self-organizing core "reasoning system" which makes use of LLMs as a knowledge base. Intrinsic reward signals - curiosity and prediction. External rewards - LLM (bootstrapping) & humans (alignment).

We're releasing a preview of OpenAI o1—a new series of AI models designed to spend more time thinking before they respond. These models can reason through complex tasks and solve harder problems than previous models in science, coding, and math. openai.com/index/introduc…