This website uses cookies
Read our Privacy policy and Terms of use for more information.
Concepts
Methods/Techniques
Models
Architectures
Infrastructure
Robotics
FMOps
The Org Age of AI
Interviews
AI 101
∇ Guide
AI builds AI
AI Agents
GenAI Unicorns
Infrastructure Unicorns
Community Twist
The History of LLMs
The History of Computer Vision
The History of World Models
Origins "who coined it"
Inference
Attention Span
Froth on the Daydream
Twitter Library
From our partners
Event Calendar
All Posts
AI Builds AI: Mapping recursive self-improvement as it happens
13 min read
Sep 16, 2026
What to change in its memory, instructions, tools, or model – and how to test whether the fix holds. Enterprise lessons from Dreamforce for your own workflows. AI Builds AI, Episode 2.
"Froth on the Daydream"
9 min read
Sep 14, 2026
RSI had quite a week! AI is already helping improve AI. Let's go deeper. What happens as more of the improvement loop itself becomes something AI can change.
7 min read
Sep 13, 2026
Nine 2026 RSI systems mapped onto seven loop stages: strategy, memory, skills, policy, weights, scaffold, evaluator. Most keep the base model frozen.
+1
17 min read
Sep 12, 2026
How models can keep learning during inference – and why test-time training could be a key to overcome the limitations of agents and world models.
1 min read
Sep 10, 2026
Without memory, agents forget what matters
AI Concepts & Techniques
11 min read
Sep 9, 2026
Triton's creator on why OpenAI is handing GPU decisions back to models with Gluon, and what still has to be true before a faster kernel counts as progress.
10 min read
Sep 8, 2026
OpenAI’s alignment questions and proposed Navier–Stokes solution, DeepMind’s genetic atlas, and new research in world models and self-improvement
Sep 4, 2026
More experience, human guidance and 8x compute all failed to make agents rethink strategy. What Metan and Recuris do instead, and where each stops short.
Sep 3, 2026
Join us September 15
Sep 1, 2026
Why Yann LeCun, Demis Hassabis, and Fei-Fei Li are converging on AI systems that represent, predict, simulate, plan, and act.
5 min read
Aug 27, 2026
In a controlled comparison, semantic code navigation reduced agent costs by 5% to 36%. It also raised a harder question: did the agent find every place that needed to change?
14 min read
Aug 24, 2026
Four years into an announced era, and I still cannot picture the thing we are running at