Intelligence.Log
2026-05-03
Extracted: 59 items. Sources: GitHub, Bluesky, X, Blogs.
grep TOPIC=
grep SOURCE=
sort --by=
GH
Opinionated Configuration Files
“This repository contains opinionated configuration files (dotfiles) for shell and development environments, likely including aliases, functions, and tool settings. It is a personal collection that may offer insights into an AI leader's workflow preferences.”
BSKY
The AI auto-reply bots from Twitter (fun fact, the software category is genuinely called "reply guy" tools) have started showing up on Bluesky now and it really, really sucks
❤️ 195 Likes|[Tooling]
BSKY
Ok so I was wondering where the hockey fans were on Bluesky, so I searched one hashtag, liked a few posts, reposted one... ... and now my "For You" feed is ONLY Habs fans talking about the game 🤣🤣🤣 And I am totally ok with this! 😁 (Discover feed most unaffected.)
❤️ 9 Likes|
BSKY
BSKY
This is great! I didn't even know #AISTATS was happening now and this is the second conference photo I have seen of it already! More please! 👍👌🙏
❤️ 2 Likes|
BSKY
BSKY
I am not sure I would agree with all of this post (by a well-known researcher at OpenAI), but the relationship between Anthropic and Claude is quite different than the relationship between other labs and their models. And that shows up in lots of ways, from models to how labs think about the future.
❤️ 137 Likes|[Safety][Agent]
BSKY
The single most accurate science fiction author writing about AI turned out to be… Douglas Adams He wrote about AIs that work best when emotionally manipulated & that guilt you in turn. And he understood there was no upper bound on test time compute for hard problem. Also dolphin communication!
❤️ 191 Likes|[Agent][Safety]
BSKY
BSKY
I'm a little late to the dunk-on-Dawkins party (but hey, let's keep the celebration rolling!) but I finally read his essay (minus the chatbot outputs he included) and I'm curious how much exposure it took for him to cook his brain so thoroughly. >>
❤️ 70 Likes|
BSKY
X
2025 LLM Year in Review
[LLM]
“DeepSeek Summary: Karpathy reflects on the key developments in LLMs over 2025, likely discussing training against auto-generated data.”
X
LLM Knowledge Bases
Something I'm finding very useful recently: using LLMs to build personal knowledge bases for various topics of research interest. In this way, a large fraction of my recent token throughput is going less into manipulating code, and more into manipulating
[LLM][RAG]
“DeepSeek Summary: Karpathy uses LLMs to construct personal knowledge bases, shifting his focus from code to knowledge management.”
X
I'm being accused of overhyping the [site everyone heard too much about today already].
[Tooling]
“DeepSeek Summary: Karpathy responds to criticism about overhyping a popular site, showing self-awareness about hype cycles.”
X
once you attach them to a good coding agent harness at least
[Agent][Tooling]
“DeepSeek Summary: LLMs become more effective when integrated into a robust coding agent framework.”
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: Effective use of coding agents requires knowing when to rely on them and when not to.”
X
Vibe coding is irresponsibly building software through dice rolls, not caring what code is produced
[Safety][Deployment]
“DeepSeek Summary: Criticizes 'vibe coding' as an irresponsible approach to software development.”
X
This may be the best guidance I've seen anywhere on writing a really good commit history.
[Tooling]
“DeepSeek Summary: Praises a resource on writing excellent commit messages.”
X
Visibility is the easiest piece. The hard part is analyzing and understanding what you're observing. I've spoken to teams recording 100k+
[Evaluation][Infra]
“DeepSeek Summary: Visibility is easy, but analyzing observations is hard; teams record 100k+ events.”
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]
“DeepSeek Summary: Agents require a sandboxed workspace to run code, install packages, and access files.”
X
as always, it's an exciting time to be working at LangChain!
[Tooling]
“DeepSeek Summary: Retweet expressing excitement about working at LangChain.”
X
The Second Pre-training Paradigm
[LLM][Fine-tuning]
“DeepSeek Summary: Jim Fan discusses a new paradigm for pre-training in AI.”
X
I've been a bit quiet on X recently. The past year has been a transformational experience.
[Agent]
“DeepSeek Summary: Jim Fan acknowledges his absence and hints at significant changes.”
X
Controversial opinion - the language best placed to win at deep learning is: F#.
[Infra]
“DeepSeek Summary: Jeremy Howard argues that F# is the best language for deep learning.”
X
Early reports from people using this are that it's the real deal. Strong coding. Good multilingual. Consistent over long contexts.
[LLM]
“DeepSeek Summary: Jeremy Howard shares positive early feedback about a new AI model or tool.”
X
reading "AI News" (previously Smol Talk) is probably the highest-leverage 45 mins
[LLM]
“DeepSeek Summary: Soumith recommends reading AI News as a high-leverage activity for 45 minutes.”
X
Sometimes we forget that NVIDIA wins because it's a software company.
[Infra]
“DeepSeek Summary: Soumith highlights NVIDIA's success is due to its software, not just hardware.”
X
Open LLMs need to get organized and co-ordinated about sharing human feedback.
[LLM][Fine-tuning]
“DeepSeek Summary: Soumith calls for better coordination among open LLM projects on human feedback sharing.”
X
MacStudio you ask? Apple Engineering's **actual** time spent on PyTorch support
[Infra][Tooling]
“DeepSeek Summary: Soumith comments on Apple Engineering's actual time spent on PyTorch support, likely in response to a question about MacStudio.”
X
Current AI is a librarian of existing knowledge. Science requires an explorer of the unknown.
[Evaluation]
“DeepSeek Summary: Chollet contrasts current AI's role as a retriever of known information with the need for AI that can explore and discover new knowledge, akin to scientific exploration.”
X
Folks who work in AI or software engineering feel like the world is changing exponential fast.
[Agent]
“DeepSeek Summary: Chollet observes that those in AI and software engineering perceive the pace of change as exponentially accelerating.”
X
Yann LeCun's Billion Dollar Bet. www.youtube.com.
[LLM][Agent][Tooling]
“DeepSeek Summary: Yann LeCun posts about a video titled 'Yann LeCun's Billion Dollar Bet', likely referencing his startup AMI Labs.”
X
Very excited to share @theworldlabs 's latest research work RTFM!! It's a real-time, ...
[Multi-modal][Infra]
“DeepSeek Summary: Fei-Fei Li announces The World Labs' real-time research work RTFM.”
X
$4.15B invested in open-source generates $8.8T of value for companies (aka $1 invested in open-source = $2,000 of value created) - Companies would need to spend 3.5 times more on software than they currently do
[LLM][Deployment][Infra]
“DeepSeek Summary: Open-source investment yields massive returns: $1 invested creates $2,000 in value for companies.”
X
X
X
On the infra side, composer 2 uses CP. This is (i think?) the first real detail from using CP on MLA. My understanding is that each rank
[Infra][LLM]
“DeepSeek Summary: Sasha Rush discusses infrastructure details about composer 2 using CP (context parallelism) on MLA (multi-head latent attention), noting it may be the first real detail from using CP on MLA.”
X
I have been compiling LLM/VLM training logbooks/chronicles. This is the one of the best sources to ...
[LLM][Fine-tuning][Infra]
“DeepSeek Summary: Stas Bekman compiles LLM/VLM training logbooks, providing a valuable resource for understanding training processes.”
X
Thanks to an awesome contribution from @omarnomad The Machine Learning Engineering Open book now can ...
[Tooling][LLM]
“DeepSeek Summary: Acknowledges a contribution to the Machine Learning Engineering Open Book, expanding its content.”
X
If you were holding off to try @MSFTDeepSpeed ZeRO++ it looks like deepspeed@master should ...
[Infra][Fine-tuning]
“DeepSeek Summary: Announces that DeepSpeed ZeRO++ is now usable, encouraging users to try it.”
X
Modern art. Artist: PyTorch memory profiler Model: Llama-8B The piece on the left is the ...
[Infra][Tooling]
“DeepSeek Summary: Uses a PyTorch memory profiler on Llama-8B to create a visual representation of memory usage.”
X
Working at Hugging Face over the past 3.5+ years has allowed me to identify what technical areas truly interest me! In turn, that has allowed me to directly...
[Fine-tuning]
“DeepSeek Summary: Reflection on how working at Hugging Face helped identify technical interests.”
X
We're shipping an elaborate guide on how to profile diffusion pipelines in Diffusers to set them...
[Deployment][Tooling]
“DeepSeek Summary: Announcement of a guide for profiling diffusion pipelines in Diffusers.”
X
I read three technical reports from Moonshot AI's Kimi K2.5 paper, Cursor's Composer 2 report and blog post, and Chroma's Context-1 write-up
[LLM][Tooling]
“DeepSeek Summary: Philipp Schmid reads and shares technical reports from Moonshot AI, Cursor, and Chroma.”
X
Told an AI agent to read the autoresearch repo and build a version for QMD. Get training data from tobi/qmd github. Went to sleep. Woke up to a 0.8B model
[Agent][Fine-tuning][Deployment]
“DeepSeek Summary: Philipp Schmid automated model training using an AI agent, resulting in a 0.8B model overnight.”
X
Random thought. We are going to be so much faster at creating and building.
[Deployment]
“DeepSeek Summary: Philipp Schmid reflects on the accelerating pace of creation and building with AI.”
X
So much work is going into faking continual learning and memory for AIs
[LLM][Evaluation]
“DeepSeek Summary: Critiques the effort spent on simulating continual learning and memory in AI systems.”
X
As someone who is pretty good at keeping up with AI, I can barely keep up with it all.
[Deployment]
“DeepSeek Summary: Even an AI expert finds the pace of AI development overwhelming.”
X
"Load bearing," "I keep coming back to," "Not X, but Y" A curse of using AI a lot is that
[LLM]
“DeepSeek Summary: Notes common phrases that emerge from heavy AI use.”
X
Ethan Mollick profile. Ethan Mollick. ✓. emollick. Apr 25. Ethan Mollick's Image on X. 32. 337. 4220. 243898 ·.
“DeepSeek Summary: Profile page, not a tweet.”
X
I work on understanding and improving training for NLP models, with a focus on studying how structures and mechanistic behaviors emerge over the
[LLM][Safety]
“DeepSeek Summary: Naomi Saphra describes her research on understanding and improving NLP model training, focusing on emergent structures and mechanistic behaviors.”
X
what a perfect space for scientific discourse! I'll start off with a few images of myself
[Evaluation]
“DeepSeek Summary: Sarcastic comment about using images of herself for scientific discourse.”
X
#throwback coz it's finally the day again!!! #HellOnWheels back on AMC 9/8c tonight!
[Deployment]
“DeepSeek Summary: Angela Zhou promotes the return of the TV show Hell on Wheels, in which she appears.”
X
Best work breaks #onset #HellonWheels -- dunno who's cuter, @ansonmount or Mac his dog?
[Deployment]
“DeepSeek Summary: Angela Zhou shares a lighthearted moment on set, comparing co-star Anson Mount to his dog.”
X
For the first time in almost a decade, I'm teaching a class on learning and control.
[Evaluation]
“DeepSeek Summary: Ben Recht is teaching a class on learning and control after nearly a decade.”
X
Building a theory of the architecture of organizing machines and people.
[Infra]
“DeepSeek Summary: Recht is developing a theory for organizing machines and people.”
X
Fully open machine learning requires not only GPU access but a community commitment to openness.
[Infra]
“DeepSeek Summary: Open ML needs both GPU access and community commitment to openness.”
BLOG
Context as infra, taste as config, verification for autonomy, scale via delegation, closing the loop.
By Eugene Yan
“The post frames working with AI as a compound process, where context serves as infrastructure, taste as configuration, and verification enables autonomy. It emphasizes scaling through delegation and closing the loop for continuous improvement. The core insight is that effective AI collaboration requires intentional design of these elements.”
-- END OF LOG --
[STATS] 59 items · Filter applied
Powered by Horizon + DeepSeek