Intelligence.Log
2026-05-07
Extracted: 57 items. Sources: GitHub, Bluesky, X, Blogs.
++ AI OVERVIEW ++
Today's discourse was anchored by a provocative reflection from Ethan Mollick, who highlighted a stark 2022 trade-off: for the cost of a single frontier AI training run ($24B), we could have had prototype vaccines ready for all 26 viral families threatening humanity. This sparked a broader conversation about global health security versus AI compute investment, a theme that resonated across the community. On the project front, several new repos trended around lightweight, local-first AI agents designed for offline task automation, signaling a shift away from cloud-dependent models. The tension between massive centralized AI spending and decentralized, socially beneficial applications remains the defining debate of the week.
grep TOPIC=
grep SOURCE=
sort --by=
GH
DeepSeek 4 Flash local inference engine for Metal
“ds4 is a local inference engine for DeepSeek 4 Flash, optimized for Apple Metal. It provides efficient on-device inference for a cutting-edge LLM, making advanced AI accessible on consumer hardware.”
BSKY
Every so often I think about how, in 2022, for $24B we could had "prototype vaccines ready for each of the 26 known viral families that cause human disease" so they can be deployed in 100 days if there was ever a need. This effort was not funded. ifp.org/why-barda-de...
❤️ 96 Likes|
BSKY
Under-reported details of the xAI/Anthropic Colossus data center deal: Anthropic get Colossus 1 but xAI keep using the larger Colossus 2, Colossus 1 has a REALLY bad environmental record, and xAI just shut down a bunch of older models on 2 weeks' notice simonwillison.net/2026/May/7/x...
❤️ 123 Likes|[Infra][Deployment]
BSKY
BSKY
@beenwrekt.bsky.social brilliant as usual: "Indeed, the language of mathematical rationality is a Bayesian language game, always working to box out the unmeasurable and unquantifiable. It demands language without ambiguity, but of course, language is always ambiguous, fluid, and evolving."
❤️ 13 Likes|[LLM][Safety]
BSKY
Visiting most of the leading Chinese AI labs, I'm struck by a culture that's extremely well suited to building LLMs with fewer resources, but one happening in a very different ecosystem, more companies at play, almost no data industry, etc. Full report: www.interconnects.ai/p/notes-from...
❤️ 83 Likes|[LLM]
BSKY
So Claude Mythos was, indeed, not marketing hype. Remember this is a general purpose model that just happens to be good at finding exploits because good models are good at lots of things. Expect similar from OpenAI & Google. And from open models in 8 months. hacks.mozilla.org/2026/05/behi...
❤️ 242 Likes|[Safety]
BSKY
@alexhanna.bsky.social and I are so excited to announce that THE AI CON has been selected as Book in Common for 2026-27 at Cal State Chico! We're excited for thousands of folks to read and engage with our work. We'll be visiting campus on April 7, 2027 for a public event.
❤️ 101 Likes|[Safety]
BSKY
Seattle friends! This event on May 17, with Shelley Fairweather-Vega at Folio: The Seattle Athenaeum should be really fun. Join us! www.folioseattle.org/event-detail...
❤️ 21 Likes|
BSKY
Are large language models mathematically rational? I swear I’m not dodging the question in this post, but it depends on your perspective.
❤️ 13 Likes|[LLM][Evaluation]
X
Drafted a blog post - Used an LLM to meticulously improve the argument over 4 hours. - LLM demolishes the entire argument and convinces me that the opposite is in fact true.
[LLM]
“DeepSeek Summary: Karpathy used an LLM to improve a blog post, but the LLM convinced him the opposite argument was true.”
X
The hottest new programming language is English
[LLM]
“DeepSeek Summary: Karpathy suggests English is becoming the new programming language due to LLMs.”
X
By training LLMs against auto-generated data, we can achieve... [content truncated in search result]
[LLM][Fine-tuning]
“DeepSeek Summary: Karpathy discusses training LLMs with auto-generated data.”
X
It's interesting how "better at code" has become the defining goal of almost every AI lab over the
[LLM][Deployment]
“DeepSeek Summary: Simon observes that AI labs are increasingly focused on improving code generation capabilities.”
X
I am not excited about visual workflow builders 1. Not simple enough for the average user
[Tooling]
“DeepSeek Summary: Harrison Chase expresses skepticism about visual workflow builders, citing lack of simplicity for average users.”
X
We launched LangSmith Agent Builder this week as a no-code way to build agents. A key part of Agent builder is it's memory system.
[Agent][Tooling]
“DeepSeek Summary: Announcement of LangSmith Agent Builder with emphasis on its memory system.”
X
In the hot path as the agent is running. The agent can decided to (or the user can prompt it to) update its memory as it is working on the core
[Agent]
“DeepSeek Summary: Discusses agent memory updates during runtime, either autonomously or via user prompt.”
X
When building agents, you need to iterate on production data much more than when building traditional software. You need to iterate on how
[Agent][Evaluation]
“DeepSeek Summary: Emphasizes the importance of iterating on production data for agent development.”
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: Argues that agents require sandboxed workspaces for code execution and file access.”
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 recent silence on X and reflects on a transformative past year.”
X
It gives me a lot of comfort knowing that we are the last generation without advanced robots everywhere.
[Agent][Multi-modal]
“DeepSeek Summary: Expresses optimism about the imminent ubiquity of advanced robots.”
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.
[Multi-modal][Agent]
“DeepSeek Summary: Defines world modeling in the context of robotics and AI.”
X
Everyone's freaking out about vibe coding. In the holiday spirit, allow me to share my anxiety on the wild
[LLM][Evaluation]
“DeepSeek Summary: Comments on the hype around 'vibe coding' and shares his own concerns.”
X
Here's a complete unedited video of asking Grok for its views on the Israel/Palestine situation. It first searches twitter for what Elon thinks.
[Safety][Evaluation]
“DeepSeek Summary: Demonstrates Grok's behavior of searching Twitter for Elon Musk's opinion on a sensitive geopolitical topic.”
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.”
X
Today, we are excited to announce Thinking Machines Lab (thinkingmachines.ai), an artificial intelligence research and product company.
[Infra]
“DeepSeek Summary: Soumith announces the launch of Thinking Machines Lab, an AI research and product company.”
X
I think it's clear that for many smaller companies that invested in deep learning, it turned out
[Deployment]
“DeepSeek Summary: Deep learning investments may not have paid off for smaller companies.”
X
Folks who work in AI or software engineering feel like the world is changing exponential fast.
[Deployment]
“DeepSeek Summary: Perception of rapid change in AI and software engineering.”
X
Current AI is a librarian of existing knowledge. Science requires an explorer of the unknown.
[Evaluation]
“DeepSeek Summary: Contrasts current AI's retrieval capabilities with the exploratory nature of science.”
X
Very excited to share @theworldlabs 's latest research work RTFM!! It's a real-time, ...
[Multi-modal][Infra][Agent]
“DeepSeek Summary: Fei-Fei Li announces World Labs' RTFM research, a real-time spatial intelligence model.”
X
So here's my postmortem after hunting for a data science job.
[Deployment]
“DeepSeek Summary: Max Woolf shares a postmortem of his data science job search.”
X
I got to cover for the excellent @HadleyFreeman in the Guardian today so
“DeepSeek Summary: Phil Wang covered for Hadley Freeman in the Guardian.”
X
Having a wonderful time hanging out with my uncle James Wong at the Chelsea Flower show!
“DeepSeek Summary: Phil Wang spent time with his uncle James Wong at the Chelsea Flower Show.”
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 discusses infrastructure details: Composer 2 uses CP (context parallelism) on MLA (Multi-head Latent Attention).”
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]
“DeepSeek Summary: Sasha humorously describes the experience of receiving a reproduction of his own paper.”
X
I have been compiling LLM/VLM training logbooks/chronicles. This is the one of the best sources to
[LLM][Infra][Fine-tuning]
“DeepSeek Summary: Stas has been compiling training logbooks for LLM/VLM, indicating a focus on documenting training processes.”
X
Thanks to an awesome contribution from @omarnomad The Machine Learning Engineering Open book now can
[Tooling][Infra]
“DeepSeek Summary: Stas acknowledges a contribution to the Machine Learning Engineering Open Book, showing community collaboration.”
X
A huge thank you note to Yih-Dar SHIEH who has been doing an amazing QA work for @huggingface for
[Tooling][Infra]
“DeepSeek Summary: Stas thanks a contributor for QA work at Hugging Face, emphasizing quality assurance in ML.”
X
This is a long overdue section of the ML Engineering Understanding Training Loss Patterns
[LLM][Fine-tuning][Infra]
“DeepSeek Summary: Stas announces a new section on understanding training loss patterns in ML Engineering.”
X
Details:
[LLM][Deployment]
“DeepSeek Summary: Sayak Paul shared a post with details, but the content is truncated.”
X
@PyTorch Of course, I forgot. Check out the docs for complete examples:
[Infra][Deployment]
“DeepSeek Summary: Sayak Paul acknowledges forgetting something and directs to PyTorch documentation.”
X
Guide: ReAct agent from scratch with Gemini 2.5 and LangGraph | Gemini API | Google AI for Developers. ai.google.dev.
[Agent][LLM]
“DeepSeek Summary: Philipp shares a guide on building a ReAct agent from scratch using Gemini 2.5 and LangGraph.”
X
How to use Deep Research with the Gemini API. www.philschmid.de.
[LLM][Tooling]
“DeepSeek Summary: Philipp explains how to use the Deep Research feature with the Gemini API.”
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
One thing thing about AI, for better and worse, is that "everything around me is somebody's life
[Safety]
“DeepSeek Summary: Reflects on the profound impact of AI on people's lives.”
X
We are starting to see some nuanced discussions of what it means to work with advanced AI In this
[Agent]
“DeepSeek Summary: Notes the emergence of nuanced discussions about working with advanced AI.”
X
Had early access to GPT-5.4 and Pro. They are very good. One fun illustration of progress,
[LLM]
“DeepSeek Summary: Early access to GPT-5.4 and Pro shows significant progress.”
X
Image is of the 1990s Microsoft writing assistant character Clippy with its eyebrows raised positioned in.
[Safety]
“DeepSeek Summary: Emily Bender posted an image of Clippy, likely to critique AI hype or anthropomorphism.”
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: Bender criticizes passive language in AI discourse, emphasizing human agency in AI outcomes.”
X
Life update: I'm starting as faculty at Boston University in 2026! BU ...
[LLM][Evaluation]
“DeepSeek Summary: Naomi Saphra announces starting as faculty at Boston University in 2026.”
X
what a perfect space for scientific discourse! I'll start off with a few images of myself
[Safety]
“DeepSeek Summary: Sarcastic comment about scientific discourse on social media.”
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][Fine-tuning]
“DeepSeek Summary: Describes her research focus on NLP model training and emergent behaviors.”
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 a long hiatus.”
X
Building a theory of the architecture of organizing machines and people.
[Infra]
“DeepSeek Summary: He is working on a theory for organizing both machines and people.”
X
How two mathematicians resolved a 50-year-old open problem by finding the solution in an 80-year-old paper
[Safety]
“DeepSeek Summary: He shares a story about mathematicians solving a long-standing problem using an old paper.”
BLOG
<p>There weren't a lot of big new announcements from Anthropic at yesterday's Code w/ Claude event, but the biggest by far was the deal they've struck with SpaceX/xAI to use "all of the capacity of their Colossus data center".</p> <p>As I mentioned in my <a...
“Anthropic has struck a deal with xAI to use the full capacity of the Colossus data center, signaling a major infrastructure collaboration. This move highlights the escalating demand for compute resources in AI development and the strategic partnerships forming to secure them.”
BLOG
Lessons from my trip to talk to most of the leading AI labs in China.
“China's AI labs are highly focused on practical applications and large-scale engineering, often prioritizing rapid iteration over theoretical novelty. The ecosystem is characterized by intense competition, strong government support, and a unique blend of open-source contributions and proprietary development.”
-- END OF LOG --
[STATS] 57 items · Filter applied
Powered by Horizon + DeepSeek