Intelligence.Log
2026-05-22
Extracted: 56 items. Sources: GitHub, Bluesky, X.
++ AI OVERVIEW ++
Today's AI landscape is being reshaped by a critical architectural breakthrough: NVlabs' Gated DeltaNet-2, which decouples the erase and write operations in linear attention, has quickly garnered attention from key researchers like Jeremy Howard. This official PyTorch implementation promises to address a fundamental limitation of current linear attention models, potentially unlocking more efficient and expressive sequence modeling. Meanwhile, in the practical tools space, the Apple TV and AirPlay client library `pyatv` saw a significant star surge, driven by a star from Max Woolf, highlighting ongoing developer interest in bridging Apple's ecosystem with open-source automation. The day's trends underscore a clear split between cutting-edge research in model efficiency and the steady demand for robust, real-world integration libraries.
grep TOPIC=
grep SOURCE=
sort --by=
GH
Official PyTorch Implementation of Gated DeltaNet-2: Decoupling Erase and Write in Linear Attention
“Gated DeltaNet-2 introduces a novel linear attention mechanism that decouples erase and write operations, improving memory management in recurrent neural networks. Built on the DeltaNet architecture, it achieves competitive performance with standard attention while maintaining linear complexity, making it suitable for long-sequence tasks.”
GH
A client library for Apple TV and AirPlay devices
“A Python client library for controlling Apple TV and AirPlay devices, enabling home automation and media streaming. It supports asyncio for async operations and is available on PyPI.”
BSKY
@futurism.com important update to your story from September 2024 futurism.com/the-byte/fac... - the FTC just punished those companies for lying about how their technology works, they weren't listening through device microphones at all
❤️ 140 Likes|
BSKY
❤️ 2 Likes|
BSKY
BSKY
I think people don't realize why Gemini Omni is different than other video AIs. It is fully multimodal, so it can edit video natively, too I took the famous "train " movie from 1896 & made it a bullet train, LEGO, added a time traveler, a centipede, muppets... (note the reflections and shadows?)
❤️ 171 Likes|
BSKY
Its funny how much the whole "strawberry" thing, which turned out to be o1-preview & Reasoners, was dismissed as overhyped at launch when it is clear in retrospect that it was way underhyped. A direct line from models unable to do basic math to solving unresolved math problems in 18 months.
❤️ 106 Likes|
BSKY
I appreciate this op-ed by my UW colleague Tomas Rocha in the Seattle Times today, calling out our state office of public education for its weak sauce "AI" policy. www.seattletimes.com/opinion/wa-s... >>
❤️ 24 Likes|
X
Personal update: I've joined Anthropic. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.
[LLM][Safety][Deployment]
“DeepSeek Summary: Karpathy announces his move to Anthropic, focusing on R&D while continuing his education passion.”
X
Excited to share that I am starting an AI+Education company called Eureka Labs.
[LLM][Deployment][Tooling]
“DeepSeek Summary: Karpathy launches Eureka Labs, an AI+Education venture.”
X
I'm being accused of overhyping the [site everyone heard too much about today already].
[LLM][Evaluation]
“DeepSeek Summary: Karpathy responds to criticism about overhyping a popular site.”
X
Vibe coding is irresponsibly building software through dice rolls, not caring what code is produced
[Agent][Safety][Tooling]
“DeepSeek Summary: Simon criticizes 'vibe coding' as an irresponsible approach to software development.”
X
A short note that the predictions that LLMs would favor 'boring technology' that's...
[LLM][Deployment]
“DeepSeek Summary: Simon notes that LLMs may not favor boring technology as predicted.”
X
I'm beginning to suspect that a key skill in working effectively with coding agents is developing an intuition for when you don't need to...
[Agent][Tooling]
“DeepSeek Summary: Simon suggests that knowing when not to use coding agents is a key skill.”
X
This seems like a good bet to me - coding agents make it no longer remotely excusable to skip out on...
[Agent][Tooling]
“DeepSeek Summary: Simon argues coding agents remove excuses for skipping certain practices.”
X
Visibility is the easiest piece. The hard part is analyzing and understanding what you're observing. I've spoken to teams recording 100k+
[Infra][Evaluation][Tooling]
“DeepSeek Summary: Harrison Chase emphasizes that visibility into systems is easy, but the real challenge is analyzing and understanding the data.”
X
TL;DR: More and more agents need a workspace: a computer where they can run code, install packages, and access files. Sandboxes provide this
[Agent][Infra][Tooling]
“DeepSeek Summary: Harrison Chase notes that AI agents increasingly require sandboxed environments to execute code and access resources.”
X
Today we're launching LangChain Labs, a new applied research effort focused on Continual Learning. Our goal is to advance open,
[LLM][Fine-tuning][Safety]
“DeepSeek Summary: Announcement of LangChain Labs, a research initiative targeting continual learning in AI.”
X
It gives me a lot of comfort knowing that we are the last generation without advanced robots everywhere.
[Agent]
“DeepSeek Summary: Reflects on the imminent proliferation of advanced robots, suggesting a future where robots become ubiquitous.”
X
Resource constraints are a beautiful thing. Survival instinct in a cut-throat AI competitive land
[LLM]
“DeepSeek Summary: Views resource constraints as a positive force driving innovation and survival in AI competition.”
X
In this context, I define world modeling as predicting the next plausible world state (or a longer duration of states) conditioned on an action.
[Agent]
“DeepSeek Summary: Defines world modeling as predicting future world states based on actions, a key concept in AI.”
X
those who think RL use less compute don't know RL at all SFT: human generates data and machine learns RL:
[Fine-tuning]
“DeepSeek Summary: Argues that reinforcement learning requires significant compute, contrary to some beliefs.”
X
reading "AI News" (previously Smol Talk) is probably the highest-leverage 45 mins
[LLM]
“DeepSeek Summary: Recommends reading 'AI News' as a high-leverage activity for staying informed.”
X
MacStudio you ask? Apple Engineering's **actual** time spent on PyTorch support
[Infra]
“DeepSeek Summary: Comments on Apple Engineering's time investment in PyTorch support.”
X
we've been working on democratizing fast kernel writing on the @PyTorch team. try
[Infra][Tooling]
“DeepSeek Summary: Announces work on democratizing fast kernel writing in PyTorch.”
X
Sometimes we forget that NVIDIA wins because it's a software company.
[Infra]
“DeepSeek Summary: Argues that NVIDIA's success is due to its software, not just hardware.”
X
It's surprisingly easy to do "hard" things -- for the most part, you need to get started and keep at it
[Agent]
“DeepSeek Summary: Chollet emphasizes that persistence and starting are key to accomplishing hard tasks.”
X
Many people assume that LRM reasoning breaks down past a certain "complexity" or "number of steps"
[LLM]
“DeepSeek Summary: Chollet challenges assumptions about limitations of language model reasoning.”
X
Reaching AGI won't be beating a benchmark. It will be the end of the human-AI gap.
[Evaluation]
“DeepSeek Summary: Chollet argues that AGI is not about benchmark performance but closing the gap between human and AI capabilities.”
X
I love Geoff. But he understands even less than Dario about the effects of technological revolutions on
[Safety]
“DeepSeek Summary: Yann LeCun critiques Geoffrey Hinton and Dario Amodei's understanding of technological revolutions.”
X
It seems to me that before "urgently figuring out how to control AI systems much smarter than us" we need
[Safety]
“DeepSeek Summary: LeCun argues that we need to address more fundamental issues before worrying about controlling superintelligent AI.”
X
The emergence of superintelligence is not going to be an event. We don't have anything close to a
[Safety]
“DeepSeek Summary: LeCun dismisses the idea of a sudden superintelligence emergence, emphasizing gradual progress.”
X
Very excited to share @theworldlabs 's latest research work RTFM!! It's a real-time,
[Multi-modal]
“DeepSeek Summary: Fei-Fei Li announces RTFM, a real-time research work from World Labs.”
X
congrats to OpenAI on winning the Turing Test
[LLM][Evaluation][Deployment]
“DeepSeek Summary: Max Woolf sarcastically congratulates OpenAI for supposedly winning the Turing Test, likely referencing ChatGPT's human-like responses.”
X
X
Been working on text feedback / OPSD in Composer. Really interesting space, and much more to be explored.
[Tooling][Agent]
“DeepSeek Summary: Sasha Rush is working on text feedback and OPSD in Composer, indicating ongoing development in AI-assisted coding tools.”
X
⛏️
[LLM]
“DeepSeek Summary: A short post with a pickaxe emoji, possibly referencing mining or digging into research.”
X
today i woke up to a living version of a phd student's nightmare. a new paper in my inbox: a detailed reproduction of a paper i wrote
[Evaluation][Safety]
“DeepSeek Summary: Sasha Rush expresses surprise and perhaps concern about a reproduction of his paper arriving in his inbox.”
X
Thanks to an awesome contribution from @omarnomad The Machine Learning Engineering Open book now can ...
[LLM][Tooling]
“DeepSeek Summary: Acknowledges contribution to the ML Engineering Open Book.”
X
This is a long overdue section of the ML Engineering Understanding Training Loss Patterns ...
[Fine-tuning][LLM]
“DeepSeek Summary: Announces a new section on training loss patterns in ML Engineering.”
X
Modern art. Artist: PyTorch memory profiler Model: Llama-8B The piece on the left is the ...
[Infra][LLM]
“DeepSeek Summary: Humorously compares PyTorch memory profiler output to modern art.”
X
MLT __init__ Paper Reading & Discussion Tim Dettmers, Mike Lewis, Younes Belkada, Luke Zettlemoyer: LLM.int8(): 8-bit Matrix Multiplication for Transformers at Scale.
[LLM][Infra][Deployment]
“DeepSeek Summary: Younes Belkada co-authored the paper on LLM.int8() for 8-bit matrix multiplication, enabling large transformers to run with reduced memory.”
X
Starting in a minute. Joining link: https://t.co/ZvFxAYMGgN
[Infra]
“DeepSeek Summary: Sayak Paul announces an upcoming event or stream with a joining link.”
X
Thanks @AnthropicAI. Thanks @huggingface for letting me work on Diffusers and other open-source projects across the fleet.
[Deployment][Infra]
“DeepSeek Summary: Sayak Paul expresses gratitude to Anthropic and Hugging Face for opportunities to work on Diffusers and other open-source projects.”
X
Excited to introduce the Gemini Interactions API, a unified interface for Gemini models and agents. Starting today with Gemini Deep Research Agent. - Unifies access to models and agents via a single RESTful endpoint.
[Agent][Infra][Tooling]
“DeepSeek Summary: Philipp Schmid announces the Gemini Interactions API, a unified RESTful interface for Gemini models and agents, starting with the Deep Research Agent.”
X
On the plus side with Opus 4.7, if it does decide to think it produces BY FAR the best
[LLM]
“DeepSeek Summary: Opus 4.7 produces the best results when it decides to think.”
X
In 1980, the philosopher John Searle proposed a thought experiment: a person locked in a room, manipulating Chinese characters according to a
[Safety][LLM]
“DeepSeek Summary: References Searle's Chinese Room thought experiment to discuss AI understanding.”
X
Talking about the ethics of AI companies or personalities, or discussing the potential of
[Safety]
“DeepSeek Summary: Discusses AI ethics and potential of AI companies.”
X
We are starting to see some nuanced discussions of what it means to work with advanced AI In this
[Agent][Deployment]
“DeepSeek Summary: Notes emerging nuanced discussions about working with advanced AI.”
X
EMILY M. BENDER: Yeah. And so passive, like, oops, the moon, the moon went further away. It's like no, actually, you made some decisions.
[Safety]
“DeepSeek Summary: Critiques passive language that deflects responsibility for AI decisions.”
X
@emilymbender.bsky.social. emilymbender. Feb 10. Image is of the 1990s Microsoft writing assistant character Clippy with its eyebrows raised positioned in.
[LLM]
“DeepSeek Summary: Uses Clippy meme to comment on AI assistants.”
X
what a perfect space for scientific discourse! I'll start off with a few images of myself
[LLM]
“DeepSeek Summary: Naomi Saphra humorously comments on a space for scientific discourse, offering to share images of herself.”
X
Life update: I'm starting as faculty at Boston University in 2026! BU has SCHEMES for LM interpretability & analysis, so I couldn't be more pumped to join a
[LLM][Evaluation]
“DeepSeek Summary: Naomi Saphra announces her upcoming faculty position at Boston University in 2026, excited about their work on language model interpretability.”
X
Very half-baked philosophy of engineering post: How do we prove that something is unpredictable in machine learning
[Safety][Evaluation]
“DeepSeek Summary: Ben Recht questions how to prove unpredictability in machine learning, reflecting on engineering philosophy.”
X
For the first time in almost a decade, I'm teaching a class on learning and control.
[Deployment]
“DeepSeek Summary: Ben Recht announces teaching a class on learning and control after a decade.”
X
I have a recommended reading list for Artificial Intelligence, and it hasn't changed since 2019.
[LLM][Evaluation]
“DeepSeek Summary: Ben Recht's AI reading list remains unchanged since 2019, suggesting foundational texts.”
-- END OF LOG --
[STATS] 56 items · Filter applied
Powered by Horizon + DeepSeek