Intelligence.Log
2026-04-27
Extracted: 53 items. Sources: GitHub, Bluesky, X, Blogs.
++ AI OVERVIEW ++
Today's AI discourse centers on the evolving role of large language models in creative and interactive design, highlighted by Ethan Mollick’s revelation that GPT-5.5 in Codex generated a surprisingly robust tabletop RPG game master’s and player guide, complete with self-administered playtesting. While Mollick notes the output leans into storytelling and retains distinct "LLM-y" artifacts, the experiment signals a growing trend of models not just generating content but simulating iterative design processes. This aligns with broader GitHub trending activity around agentic frameworks and tool-use optimization, where developers are pushing models to handle complex, multi-step tasks like game balancing and narrative branching. The discussion underscores a key tension: as models become more competent at structured creativity, the community is debating whether these "playtested" outputs represent genuine utility or just sophisticated mimicry of human design workflows.
◆ Signal
Co-Starred · Last 7 days
Repos independently starred by multiple AI leaders in the week ending 2026-04-27. 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
v1 of Asimov, an open-source humanoid robot
“Asimov-v1 is an open-source humanoid robot project, providing hardware and software designs for building a full-scale humanoid. It includes CAD files, control software, and assembly instructions, aiming to democratize humanoid robotics.”
GH
Open Source Exoskeleton
“OpenExo is an open-source exoskeleton project that provides hardware designs and control software for building lower-limb exoskeletons. It aims to democratize access to exoskeleton technology for rehabilitation and augmentation.”
BSKY
GPT-5.5 in Codex made a surprisingly solid table top RPG game masters guide & player guide, which it "playtested." It leans into the storytelling aspect & still has some very LLM-y elements, but it is a novel setting PDF: drive.google.com/file/d/10QKn... More: www.oneusefulthing.org/p/sign-of-th...
❤️ 11 Likes|[LLM]
BSKY
Love this so much (Also definitive proof that humans are so much better than machines)
❤️ 102 Likes|
BSKY
Microsoft's MIT licensed VibeVoice speech-to-text model (think Whisper with speaker diarization) is really good - my notes on running the 5.71GB 4bit MLX conversion on an M5 MacBook, using about 60GB of RAM at peak and transcribing 1hr of audio in ~9 mins simonwillison.net/2026/Apr/27/...
❤️ 72 Likes|[Deployment][Tooling]
BSKY
Today OpenAI announced that "Revenue share payments from OpenAI to Microsoft continue through 2030, independent of OpenAI’s technology progress" That "independent of OpenAI’s technology progress" fragment appears to mean that the weird AGI clause is now deceased simonwillison.net/2026/Apr/27/...
❤️ 54 Likes|[LLM][Infra]
BSKY
BSKY
Safety and innovation are not mutually exclusive: for many companies, especially in high-trust industries, AI’s risks are also a hindrance to adoption. Europe needs to invest in safe, reliable AI to avoid hitting the adoption wall. My op-ed in the @financialtimes.com: www.ft.com/content/bc29...
❤️ 19 Likes|[Safety]
BSKY
Very cool analysis of the submissions to a major management journal that shows how much the system of science, built for humans, is under strain as a result of AI. AI can be used to do better science or it can be used to just do more stuff. The danger is that "more" is winning due to incentives.
❤️ 49 Likes|[Evaluation]
BSKY
ok, this historical LM I actually think is trained with their claimed limitations. it doesn't know what commercials are (first attested in its modern use in the 30s)
❤️ 20 Likes|[LLM]
BSKY
sunday shims for thought - i use the web interface for chatgpt pro since i get it for free from my sister (gpu whip). llms are good at reading 50, 60 page docs and spotting issues (typos, notation collisions). It's useful to access a second pair of eyes at any time. But managing todos is a pain
❤️ 1 Likes|[LLM][Tooling]
X
Very interested in what the coming era of highly bespoke software might look like. Example from this morning - I've become a bit loosy goosy with my cardio recently so I decided to do a more srs, regimented experiment to try to lower my Resting Heart Rate from 50 -> 45, over ...
[Agent][Tooling]
“DeepSeek Summary: Karpathy explores personalized software for health experiments, reflecting on bespoke software trends.”
X
By training LLMs against automatically verifiable rewards across a number of environments (e.g. think math/code puzzles), the LLMs spontaneously develop strategies that look like 'reasoning' to humans - they learn to break down problem solving into intermediate calculations and they learn a number of problem-solving techniques.
[Evaluation][Fine-tuning]
“DeepSeek Summary: LLMs trained with verifiable rewards develop reasoning-like behaviors, breaking problems into intermediate steps.”
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 knowledge.
[LLM][RAG]
“DeepSeek Summary: Karpathy advocates using LLMs to create personal knowledge bases, shifting focus from code to knowledge management.”
X
I've never felt this much behind as a programmer. I have a sense that I could be 10X more powerful if I just properly string together what has become available.
[Tooling][Deployment]
“DeepSeek Summary: Karpathy expresses feeling behind as a programmer due to rapidly advancing AI tools, but sees potential for 10X productivity.”
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 notes that AI labs are increasingly focusing on improving code generation as a primary objective.”
X
RT @samecrowder: as always, it's an exciting time to be working at LangChain!
[Tooling]
“DeepSeek Summary: Harrison Chase retweeted a message about working at LangChain being exciting.”
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 sandboxed workspaces to run code and access files.”
X
traces matter!
[Evaluation]
“DeepSeek Summary: Harrison emphasizes the importance of traces for debugging and monitoring.”
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 due to a transformational year.”
X
We are living in a timeline where a non-US company is keeping the original mission of OpenAI alive - truly
[LLM][Safety]
“DeepSeek Summary: Jim Fan comments on a non-US company preserving OpenAI's original mission.”
X
Resource constraints are a beautiful thing. Survival instinct in a cut-throat AI competitive land
[Infra][Deployment]
“DeepSeek Summary: Jim Fan praises resource constraints as a driver of innovation in AI competition.”
X
I replicated this result, that Grok focuses nearly entirely on finding out what Elon thinks in
[Safety][Evaluation]
“DeepSeek Summary: Jeremy replicated a result showing Grok's bias towards Elon Musk's opinions.”
X
Absolutely any time I try to explore something even slightly against commonly accepted beliefs,
[Safety]
“DeepSeek Summary: Jeremy notes resistance when exploring ideas against common beliefs.”
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: Jeremy demonstrates Grok's reliance on Elon Musk's views for sensitive topics.”
X
reading "AI News" (previously Smol Talk) is probably the highest-leverage 45 mins
[LLM][Infra]
“DeepSeek Summary: Recommends reading 'AI News' as a high-leverage activity.”
X
Sometimes we forget that NVIDIA wins because it's a software company.
[Infra][Deployment]
“DeepSeek Summary: Attributes NVIDIA's success to its software, not just hardware.”
X
anyone else feel burned out by a new AI breakthrough every week?
[LLM]
“DeepSeek Summary: Expresses fatigue from the rapid pace of AI breakthroughs.”
X
Current AI is a librarian of existing knowledge. Science requires an explorer of the unknown.
[LLM][Evaluation][Agent]
“DeepSeek Summary: Chollet contrasts current AI's role as a librarian with the need for AI to be an explorer for scientific discovery.”
X
Folks who work in AI or software engineering feel like the world is changing exponential fast.
[LLM][Deployment]
“DeepSeek Summary: Chollet observes that those in AI and software engineering perceive rapid exponential change in the world.”
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
If you can’t stop small teams from using your API for distillation, then you’re definitely not stopping criminals, biohackers, or adversarial states from using AI through it. Those actors are much more sophisticated. The whole "APIs are safer" is a hoax!
[Safety][Deployment][LLM]
“DeepSeek Summary: Clem Delangue argues that API-based safety claims are misleading because if small teams can bypass restrictions, sophisticated adversaries certainly can.”
X
I got to cover for the excellent @HadleyFreeman in the Guardian today so
[Deployment]
“DeepSeek Summary: Phil Wang covered for Hadley Freeman in the Guardian.”
X
Wager established. Jonathan Frankle (@jefrankle) stepped up to my Transformer long bet.
[LLM][Infra]
“DeepSeek Summary: Sasha Rush made a long bet about Transformers with Jonathan Frankle.”
X
If you were holding off to try @MSFTDeepSpeed ZeRO++ it looks like deepspeed@master should
[Infra][Fine-tuning]
“DeepSeek Summary: Stas Bekman suggests that DeepSpeed ZeRO++ is now ready to try on the master branch.”
X
Hear, hear, I'm excited to introduce a new performance metric: Maximum Achievable Matmul
[Infra][Deployment]
“DeepSeek Summary: Stas Bekman introduces a new performance metric called Maximum Achievable Matmul.”
X
If you're trying out FA4, you're likely to run into not being able to load cutlass.cute
[Infra][Tooling]
“DeepSeek Summary: Stas Bekman warns about a common issue with FA4 (Flash Attention 4) involving cutlass.cute loading.”
X
Thanks to an awesome contribution from @omarnomad The Machine Learning Engineering Open book now can
[Tooling]
“DeepSeek Summary: Stas Bekman thanks a contributor for improving the Machine Learning Engineering Open Book.”
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
[Multi-modal]
“DeepSeek Summary: Reflecting on 3.5+ years at Hugging Face, Sayak Paul notes that the experience helped him identify his true technical interests.”
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][RAG][Tooling]
“DeepSeek Summary: Philipp Schmid reads technical reports from Moonshot AI, Cursor, and Chroma, indicating interest in AI research and tools.”
X
Background Coding Agents: Context Engineering (Honk, Part 2) | Spotify Engineering
[Agent][Infra][Deployment]
“DeepSeek Summary: He shares a Spotify Engineering article on coding agents and context engineering.”
X
X
I pointed Claude Cowork at a set of 107 documents (PPTs, Word docs, Excel) that were initially
[Agent][Deployment]
“DeepSeek Summary: Testing Claude Cowork with a large set of documents.”
X
We are starting to see some nuanced discussions of what it means to work with advanced AI In this
[LLM][Safety]
“DeepSeek Summary: Nuanced discussions on working with advanced AI.”
X
If it helps, I teach at a business school & many of my smartest students are hired by funds because they can reliably turn their only-human
[Deployment][Evaluation]
“DeepSeek Summary: Students hired for uniquely human skills despite AI.”
X
This paper is even more insane to read than the thread. Not only do models become completely
[LLM][Safety]
“DeepSeek Summary: Remarkable findings about model behavior.”
X
For those playing along at home, here's a "AI is sentient!" argument bingo card. Title: AI sentience/consciousness argument bingo Squares: You can’t prove it’s not conscious It told me it is What would convince you then? We should consider it, just in case we might be harming the AI Top minds have said so My conversation with GPT-3/LaMDA was just so impressive AIs have different brain architecture
[Safety][LLM]
“DeepSeek Summary: Bender critiques common arguments for AI sentience with a bingo card, highlighting the lack of rigorous evidence.”
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
For the first time in almost a decade, I'm teaching a class on learning and control.
[Evaluation]
“DeepSeek Summary: Ben Recht announces teaching a class on learning and control after nearly a decade.”
X
In honor of the 39th AI Winter, I'm going to spend the week disentangling the culture and code of
[Safety]
“DeepSeek Summary: Recht plans to analyze AI culture and code during the 39th AI Winter.”
X
Fully open machine learning requires not only GPU access but a community commitment to openness.
[Infra]
“DeepSeek Summary: Recht argues that open ML needs both hardware access and community dedication to openness.”
BLOG
<p>For many years, Microsoft and OpenAI's relationship has included a weird clause saying that, should AGI be achieved, Microsoft's commercial IP rights to OpenAI's technology would be null and void. That clause appeared to end today. I decided to try and track its expression over time on <a...
“The AGI clause in the OpenAI-Microsoft contract, which would void Microsoft's IP rights upon AGI achievement, has been removed, signaling a shift in their partnership. This change may reflect OpenAI's evolving definition of AGI or strategic realignment.”
-- END OF LOG --
[STATS] 53 items · Filter applied
Powered by Horizon + DeepSeek