Feedforward
Personal blogs from the AI community
- Simon Willison's Weblogsimonwillison.net
May 2026 newsletter
- Understanding AIunderstandingai.org
OpenAI’s math breakthrough played to AI’s strengths
- One Useful Thingoneusefulthing.org
Choosing to Stay Human
- Interconnectsinterconnects.ai
Some ideas for what comes next, May 2026
- Posts on Max Woolf's Blogminimaxir.com
The mysterious Hy3 LLM is topping OpenRouter Model Rankings by a large margin
- Lossfunk Lettersletters.lossfunk.com
How to make progress on metaphysical puzzles in AI?
- Shreya Shankarsh-reya.com
Exploring Agent-Assisted Qualitative Analysis
- Sebastian Raschka, PhDsebastianraschka.com
Recent Developments in LLM Architectures: KV Sharing, mHC, and Compressed Attention
- Ponder on Franz Louis Cesistaleloykun.github.io
Lean4-TileLang Tensor Program Superoptimizer [WIP]
- Sander Dielemansander.ai
Learning the integral of a diffusion model
- karpathykarpathy.bearblog.dev
Sequoia Ascent 2026 summary
- Maharshi's blogmaharshi.bearblog.dev
Identity and ArithTuple Tensors in CuTeDSL
- Sam Altmanblog.samaltman.com
-
- Adjacent Possibleadjacentpossible.substack.com
The Legible Society
- Exploring Language Modelsnewsletter.maartengrootendorst.com
A Visual Guide to Gemma 4
- Chris McCormickmccormickml.com
Optimizing Training with FlashAttention varlen
- flurries of latent creativityblog.singleton.io
Cheap turpentine
- zhengdongwang.comzhengdongwang.com
The means of some change
- Token for Tokenblog.jxmo.io
How to train the best embedding model in the world
- Nicholas Carlininicholas.carlini.com
How to win a best paper award
- Andrej Karpathy blogkarpathy.github.io
microgpt
- AI: A Guide for Thinking Humansaiguide.substack.com
On Brian Cantwell Smith and the Promise of AI
- antifragile systemsyongzx.substack.com
CoT monitorability: why g-means and not F1?
- Victoria Krakovnavkrakovna.wordpress.com
2025-26 New Year review
- Yi Tayyitay.net
2025 introspections: my year back at Google
- Nick’s Substacknickjiang.substack.com
Teaching Algorithms in Ethiopia
- Neel Nandaneelnanda.io
MATS Applications Open (Due Aug 29)
- Blog - Jason Weijasonwei.net
Life lessons from reinforcement learning
- atharva's blogksagar.bearblog.dev
how we accidentally solved robotics by watching 1 million hours of YouTube
- Lil'Loglilianweng.github.io
Why We Think
- jeremybernste.injeremybernste.in
Deriving Muon
- Thonk From First Principlesthonking.ai
Why PyTorch is an amazing place to work... and Why I'm Joining Thinking Machines
- Jeremy Jordanjeremyjordan.me
Training extremely large neural networks across thousands of GPUs.
- Chip Huyenhuyenchip.com
Common pitfalls when building generative AI applications
- FOR OUR POSTERITYforourposterity.com
SITUATIONAL AWARENESS: The Decade Ahead
- ruder.ioruder.io
The Evolving Landscape of LLM Evaluation
- Jascha’s blogsohl-dickstein.github.io
Neural network training makes beautiful fractals
- Kemal Erdem Blog RSS Feederdem.pl
Step by Step visual introduction to Diffusion Models.
- Brian Kitanoblog.briankitano.com
Llama from scratch (or how to implement a paper without crying)
- siboehmsiboehm.com
Can Function Inlining Affect Floating Point Outputs? Exploring FMA and Other Consistency Issues
- peterbloem.nlpeterbloem.nl
What design can teach us
- Greg Brockmanblog.gregbrockman.com
It's time to become an ML engineer