Subscribe
Sign in
Home
Podcast
Archive
About
Latest
Top
Discussions
How Astra Edited Qwen’s Weights
From diagnosing repetitive reasoning to a persistent patch, and what the results establish about weight-space programming.
Nov 9
•
Damien Benveniste
6
How LLM Distillation Trains Smaller Models
From teacher-written answers to probability targets and feedback on the student’s own output.
Nov 6
•
Damien Benveniste
14
3
How Quantization Compresses LLMs
Follow GPTQ, AWQ, and SmoothQuant through weight compensation, channel scaling, and low-precision execution.
Nov 2
•
Damien Benveniste
12
1
October 2026
How CUDA Graphs Speed Up LLM Inference
Replay GPU operations with fresh tokens, growing caches, and changing batches.
Oct 30
•
Damien Benveniste
12
How LLM Rerankers Order Retrieved Passages
Follow pointwise scores, pairwise preferences, and listwise windows from model judgment to useful context.
Oct 26
•
Damien Benveniste
7
How Looped Transformers Reuse Their Weights
Follow changing attention states, shared gradients, and the real cost of extra passes.
Oct 23
•
Damien Benveniste
9
1
How the KV Cache Shapes AI Agent Costs
Why shrinking the cache, reusing a prefix, and keeping it nearby save different resources.
Oct 19
•
Damien Benveniste
5
1
How Hybrid LLMs Combine Memory, Attention, and Experts
Follow Mamba-2, Gated DeltaNet and KDA from state updates to the costs that active parameter counts miss.
Oct 16
•
Damien Benveniste
8
1
How Persistent Agents Keep Working
Reconstruct a coding task’s context, preserve accepted work, and recover safely when a worker disappears.
Oct 12
•
Damien Benveniste
8
1
How Self-Improving Agents Learn from Experience
How executable tasks and adaptive curricula turn production failures into useful practice.
Oct 9
•
Damien Benveniste
6
How DeepSeek V4.1 Flash Separates Reading and Generation
How a shorter input path, shared attention memory, and selective reconstruction change the cost of long conversations.
Oct 7
•
Damien Benveniste
4
1
Build a Jev-Supervised AI Agent
Typed decisions, guarded tools, and a measured comparison with LLM supervisors
Oct 3
•
Damien Benveniste
8
2
This site requires JavaScript to run correctly. Please
turn on JavaScript
or unblock scripts
