Intelligence.Log
2026-04-23
Extracted: 68 items. Sources: GitHub, Bluesky, X, Blogs.
++ AI OVERVIEW ++
Today's focus is on practical AI tools and emerging risks. The open-source community is buzzing about **run-llama/liteparse**, a fast document parser that gained significant traction after being starred by developer Simon Willison. Meanwhile, AI safety is a key discussion point, with Thomas Dietterich highlighting sophisticated new phishing techniques enabled by generative models. On a creative note, Ethan Mollick showcased the impressive stylistic capabilities of a new image model, GPT-imagegen-2, which can seamlessly blend objects like shoes with iconic artistic masterpieces.
◆ Signal
Co-Starred · Last 7 days
Repos independently starred by multiple AI leaders in the week ending 2026-04-23. Stronger signal = more overlap.
huggingface/ml-intern
×2 starrers▲ 7/10★ 688
🤗 ml-intern: an open-source ML engineer that reads papers, trains models, and ships ML models
grep TOPIC=
grep SOURCE=
sort --by=
GH
A fast, helpful, and open-source document parser
“Liteparse is a fast, open-source document parser that efficiently extracts text from PDFs and images using OCR. It provides a helpful API for developers needing reliable document processing with TypeScript support.”
GH
Create a VirtualBox machine with Pelias geocoder
“Pelias is an open-source geocoder that uses OpenStreetMap data. This Vagrant setup provides a quick way to run Pelias locally in a VirtualBox machine, simplifying deployment for development or testing.”
BSKY
BSKY
GPT-imagegen-2: "a gallery of shoes, where each shoe is under a painting & is styled matched to that painting: Starry Night, The Bathers, The Girl with the Pearl Earring, The Bayeux Tapestry, Klint's Grupp Svanen nr 17, Kandinsky's Swinging, The Garden of Earthly Delights" "now the full outfits"
❤️ 4 Likes|[Multi-modal]
BSKY
I vibe coded a web UI to LiteParse by LlamaIndex, a neat open source library for extracting text from annoying layouts PDFs simonw.github.io/liteparse/ More about how I built it on my blog: simonwillison.net/2026/Apr/23/...
❤️ 46 Likes|[Tooling]
BSKY
GPT-5.5 is out but not yet in the OpenAI API - but in the ongoing battle with Anthropic over OpenClaw OpenAI made their Codex tool's private backend API a semi-approved mechanism... and that one has GPT-5.5... So I used it to draw some pelicans simonwillison.net/2026/Apr/23/...
❤️ 53 Likes|[LLM][Deployment]
BSKY
BSKY
Ok, look at the time--- it's April 23rd!!! THE DAY YOU WILL POST ALL YOUR PHOTOS AND VIDEOS OF PAPERS, PRESENTATIONS, AND SOCIALS HAPPENING AT @iclr-conf.bsky.social #iclr #iclr2026 ON BLUESKY! Bring that conference experience to your favorite social network.. we all benefit! 😃🙏
❤️ 10 Likes|
BSKY
Some thoughts on why I don’t think banning Chinese models over distillation is worth it. open.substack.com/pub/natolamb...
❤️ 6 Likes|[Safety][Evaluation]
BSKY
Great podcast from someone building LLMs within China, on how the model building is done and thought about there. podcasts.apple.com/us/podcast/a...
❤️ 1 Likes|[LLM][Fine-tuning]
BSKY
I’m in Madrid this week for the first in-person meeting of the United Nations Independent International Scientific Panel on AI. I'm thrilled to be part of this distinguished group of 40 international experts appointed to provide an evidence-based scientific assessment of the state of AI.
❤️ 14 Likes|[Safety]
BSKY
Among all the model release furor, important to note that people don’t need to switch providers or declare a winner every time a new model is released. Especially as Opus 4.7 is a good model too! (Especially since adaptive thinking got better)
❤️ 29 Likes|[LLM]
BSKY
Here’s my view on GPT-5.5, which I have been testing for a couple of weeks. It conducted not-bad social science research on its own, developed a novel RPG & more. There is still jaggedness but GPT-5.5 Pro is (for today) the best model for hard problems. open.substack.com/pub/oneusefu...
❤️ 54 Likes|[LLM][Agent][Evaluation]
BSKY
I just submitted a paper that it took me three years to expand from the original conf submission and it feels like I wrenched another being out of my body and on the way out it turned myself inside out
❤️ 11 Likes|
BSKY
Assembling a reading list for a class on the theory of engineering architecture.
❤️ 16 Likes|[Infra]
X
Image post about training large deep neural networks
[LLM][Infra]
“DeepSeek Summary: Visual content related to neural network training, likely showing technical work or results.”
X
The hottest new programming language is English
[LLM][Tooling]
“DeepSeek Summary: Suggests that natural language (English) is becoming a primary interface for programming and AI interaction.”
X
LLM Knowledge Bases Something I'm finding very useful recently: using LLMs to build personal knowledge bases for various topics of research interest. Data ingest: I index source documents (articles, ...)
[LLM][RAG][Tooling]
“DeepSeek Summary: Describes using LLMs to create personal knowledge bases by indexing and processing source documents for research.”
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: Working with coding agents requires developing intuition about when not to use them, suggesting efficiency comes from selective application rather than constant reliance.”
X
Vibe coding is irresponsibly building software through dice rolls, not caring what code is produced
[Agent][Safety]
“DeepSeek Summary: Criticizes 'vibe coding' as irresponsible software development that relies on random outcomes without concern for code quality or correctness.”
X
This means that operations you would do on code in the software world, you now do on traces in the agent world. Debugging, testing, profiling
[Agent][Tooling][Evaluation]
“DeepSeek Summary: Harrison Chase draws a parallel between software engineering practices and agent development, emphasizing the importance of traces for debugging, testing, and profiling in the agent world.”
X
traces matter!
[Agent][Tooling]
“DeepSeek Summary: Harrison Chase succinctly emphasizes the critical importance of traces in the development and operation of AI agent systems.”
X
The first time I met Jensen was also the first time I met @elonmusk. I was interning at OpenAI that day and
[Agent]
“DeepSeek Summary: Jim Fan recalls meeting both Jensen Huang and Elon Musk for the first time during his OpenAI internship.”
X
Resource constraints are a beautiful thing. Survival instinct in a cut-throat AI competitive land
[Infra][Deployment]
“DeepSeek Summary: Jim Fan views resource limitations positively as a driver of innovation and survival in the competitive AI field.”
X
I've been a bit quiet on X recently. The past year has been a transformational experience.
[Agent]
“DeepSeek Summary: Jim Fan acknowledges a period of reduced social media activity, describing the past year as personally transformative.”
X
It gives me a lot of comfort knowing that we are the last generation without advanced robots everywhere.
[Safety][Deployment]
“DeepSeek Summary: Jim Fan finds solace in the idea that current humans are the final generation to live before ubiquitous advanced robotics.”
X
Everyone's freaking out about vibe coding. In the holiday spirit, allow me to share my anxiety on the wild
[Tooling][LLM]
“DeepSeek Summary: Jim Fan comments on the widespread concern over 'vibe coding' and intends to share his own related anxieties.”
X
Something that drives me to distraction in discussion of AI alignment: someone will say 'Oh, it's crucial we build systems with properties X'
[Safety]
“DeepSeek Summary: Jeremy Howard expresses frustration with superficial discussions about AI alignment, where people emphasize building systems with certain properties without deeper engagement.”
X
I can't begin to describe how life-changing this new project, ShellSage, has been for me over the last few weeks.
[Tooling]
“DeepSeek Summary: Jeremy Howard shares his enthusiasm for a new project called ShellSage, describing it as transformative in his recent experience.”
X
reading 'AI News' (previously Smol Talk) is probably the highest-leverage 45 mins
[LLM]
“DeepSeek Summary: Soumith Chintala highlights the value of staying informed through AI-focused content, suggesting it's a highly efficient use of time.”
X
Sometimes we forget that NVIDIA wins because it's a software company.
[Infra][Tooling]
“DeepSeek Summary: Chintala points out that NVIDIA's success is rooted in its software capabilities, not just hardware, challenging common perceptions.”
X
MacStudio you ask? Apple Engineering's actual time spent on PyTorch support
[Infra][Tooling]
“DeepSeek Summary: Chintala comments on Apple's engineering efforts regarding PyTorch support, possibly in response to questions about MacStudio.”
X
Folks who work in AI or software engineering feel like the world is changing exponential fast.
[Agent]
“DeepSeek Summary: AI and software engineering professionals perceive the world as changing at an exponential pace.”
X
I think it's clear that for many smaller companies that invested in deep learning, it turned out...
[Deployment][Evaluation]
“DeepSeek Summary: Suggests that deep learning investments had mixed or unexpected outcomes for smaller companies.”
X
Very excited to share @theworldlabs's latest research work RTFM!! It's a real-time...
[Multi-modal][Deployment]
“DeepSeek Summary: Announcing RTFM research from World Labs, focusing on real-time capabilities.”
X
AI’s next frontier is Spatial Intelligence, a technology that will turn seeing into reasoning, perception into action, and imagination into creation. This is an insanely large world created using our 3D world generation model.
[Multi-modal][Agent]
“DeepSeek Summary: Defines Spatial Intelligence as AI's next frontier, enabling reasoning and action from perception.”
X
🤗, the open and collaborative platform for AI builders. Great research on open-source by: - $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
[Infra][Tooling]
“DeepSeek Summary: Delangue highlights the immense economic value of open-source investment, showing that $1 invested generates $2,000 in value and companies would need to spend 3.5x more without it.”
X
X
Attention and RNNs are distinct, but Linear Attention is a Linear RNN, but not vice versa.
[LLM]
“DeepSeek Summary: Clarifies the relationship between attention mechanisms and RNNs, noting that linear attention can be viewed as a linear RNN, but the reverse is not true.”
X
I have been compiling LLM/VLM training logbooks/chronicles. This is the one of the best sources to...
[LLM][Multi-modal][Tooling]
“DeepSeek Summary: Stas Bekman is compiling training logbooks/chronicles for LLMs and VLMs, suggesting he's documenting best practices and insights from model training processes.”
X
Thanks to an awesome contribution from @omarnomad The Machine Learning Engineering Open book now can...
[Tooling][Deployment][Infra]
“DeepSeek Summary: Stas Bekman acknowledges a contributor to the Machine Learning Engineering Open book, indicating his involvement in creating educational resources for ML engineering.”
X
To remind - this is the memory saving you get when enabling TiledMLP :) Left: normal memory...
[Infra][Tooling][Deployment]
“DeepSeek Summary: Stas Bekman demonstrates memory savings achieved by enabling TiledMLP, showing practical optimization techniques for ML systems.”
X
Modern art. Artist: PyTorch memory profiler Model: Llama-8B The piece on the left is the...
[LLM][Tooling][Infra]
“DeepSeek Summary: Stas Bekman humorously compares PyTorch memory profiling visualizations to modern art while working with Llama-8B model.”
X
Giving a talk here is by far the most fulfilling experience of my life!
[Deployment]
“DeepSeek Summary: Sayak Paul expresses that giving a talk was the most fulfilling experience of his life, indicating significant personal and professional satisfaction from public speaking engagements.”
X
My model is a hipster doing 'it's a feature, not a bug' round.
[Evaluation]
“DeepSeek Summary: A humorous, metaphorical tweet comparing a machine learning model's unexpected behavior to a hipster's quirky justification, highlighting the often unpredictable nature of model outputs.”
X
Philipp Schmid profile. Philipp Schmid. ✓ Affiliate badge. _philschmid. 6h. _philschmid
“DeepSeek Summary: Profile update or activity indicator showing recent online presence (6 hours ago).”
X
Philipp Schmid (@_philschmid). 83 likes 3 replies ... Philipp Schmid profile. Philipp Schmid. ✓ Affiliate badge.
“DeepSeek Summary: A post that received significant engagement with 83 likes and 3 replies.”
X
Philipp Schmid (@_philschmid). 14 likes. Blog ... Philipp Schmid profile. Philipp Schmid. ✓ Affiliate badge. _philschmid.
“DeepSeek Summary: Post linking to or mentioning a blog, with moderate engagement (14 likes).”
X
I pointed Claude Cowork at a set of 107 documents (PPTs, Word docs, Excel) that were initially...
[Agent][Tooling]
“DeepSeek Summary: Testing Claude Cowork's ability to process and analyze a large collection of mixed-format documents (107 PPTs, Word docs, Excel files).”
X
My most popular AI post was a bunch of made-up 'graphs' four years ago.
[Evaluation]
“DeepSeek Summary: Reflecting on how early, speculative AI content (including fabricated visualizations) gained significant traction compared to current, more substantive discussions.”
X
We are starting to see some nuanced discussions of what it means to work with advanced AI...
[Deployment][Safety]
“DeepSeek Summary: Observing a shift towards more sophisticated and detailed conversations about the practical implications and challenges of collaborating with powerful AI systems.”
X
Teaching an experimental class for MBAs on 'vibefounding,' the students have four days to come up and...
[Deployment]
“DeepSeek Summary: Running a hands-on MBA course focused on 'vibefounding'—rapid, vibe-driven startup ideation and validation within a tight four-day timeframe.”
X
Classic study gave 146 economist teams the same dataset & got wildly different answers...
[Evaluation]
“DeepSeek Summary: Citing research showing significant variability in conclusions drawn by different expert teams analyzing identical data, underscoring subjectivity in analysis.”
X
Image is of the 1990s Microsoft writing assistant character Clippy with its eyebrows raised positioned in.
[Agent][Tooling]
“DeepSeek Summary: Emily M. Bender posted an image of the Microsoft Clippy character with raised eyebrows, likely making a commentary on AI assistants or anthropomorphism in technology.”
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][Deployment]
“DeepSeek Summary: Bender critiques passive language around AI failures, emphasizing that outcomes result from deliberate decisions rather than accidents.”
X
Look what @alexhanna and I got to do! (Hang out with the cool kids ... We're talking about the Turing Test, the grandmother of all tests for AI sentience. Joining us are AI researchers Alex Hanna and Emily M. Bender
[Evaluation][LLM]
“DeepSeek Summary: Bender references discussing the Turing Test with Alex Hanna, framing it as a foundational but potentially flawed benchmark for AI sentience.”
X
This book starts like it's gonna be a fun microhistory of TB (it gave us the Stetson!)
“DeepSeek Summary: Naomi Saphra comments on a book about tuberculosis history, noting its engaging start and a surprising connection to the Stetson hat.”
X
New preprint! Everyone loves causal interp. It's coherently defined! It makes testable predictions
[LLM][Evaluation]
“DeepSeek Summary: Announces a new preprint on causal interpretation in AI, emphasizing its clear definition and ability to produce testable predictions.”
X
Waiting on a robot body. All opinions are universal and held by both employers and family. Now a dedicated grok hate account. Accepting ML/NLP PhD students.
[LLM][Safety]
“DeepSeek Summary: Describes her current status humorously, including waiting for technology, universal opinions, opposition to 'grok', and openness to PhD applicants in ML/NLP.”
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: Announces her upcoming faculty position at Boston University in 2026, excited about their programs for language model interpretability and analysis.”
X
#throwback coz it's finally the day again!!! #HellOnWheels back on AMC 9/8c tonight!
“DeepSeek Summary: Angela Zhou expresses excitement about the return of the TV show 'Hell on Wheels' to AMC.”
X
In the first applications lecture of the semester, we build a general framework for generating the real
[Evaluation]
“DeepSeek Summary: Teaching a framework for inverse problems in applications lecture. Focus on generating real-world solutions from data.”
X
For the first time in almost a decade, I'm teaching a class on learning and control.
[Evaluation]
“DeepSeek Summary: Returning to teaching a course on learning and control after nearly ten years. Indicates renewed focus on this intersection.”
X
Everyone knows actions are fundamentally different than predictions, but it's hard to write this
[Agent][Deployment]
“DeepSeek Summary: Acknowledges the conceptual gap between prediction and action in AI/ML systems. Suggests this distinction is important but challenging to formalize.”
BLOG
<p>LlamaIndex have a most excellent open source project called <a href="https://github.com/run-llama/liteparse">LiteParse</a>, which provides a Node.js CLI tool for extracting text from PDFs. I got a version of LiteParse working entirely in the browser, using most of the same libraries that...
“LiteParse is a Node.js CLI tool for extracting text from PDFs, and this post shows how to run it entirely in the browser using the same libraries. The key insight is that many server-side tools can be adapted for client-side execution, enabling new interactive applications without backend dependencies.”
BLOG
<p><a href="https://openai.com/index/introducing-gpt-5-5/">GPT-5.5 is out</a>. It's available in OpenAI Codex and is rolling out to paid ChatGPT subscribers. I've had some preview access and found it to be a fast, effective and highly capable model. As is usually the case these days, it's hard to...
“GPT-5.5 is now available via OpenAI Codex and rolling out to paid ChatGPT users. The author finds it fast, effective, and highly capable, with improvements in coding and reasoning tasks.”
-- END OF LOG --
[STATS] 68 items · Filter applied
Powered by Horizon + DeepSeek