memory-models Podcast Summaries
memory-models on Yedapo: 2 summarized podcast and YouTube episodes. Each includes key takeaways, core concepts and notable quotes with timestamps.
Why Multi-Threaded Code Can Sometimes Misbehave (Weak Memory Concurrency) - Computerphile
Computerphile
May 18, 2026
Modern hardware and compilers prioritize performance over intuitive execution, leading to 'weak memory' behaviors where instructions are reordered or delayed in store buffers. While high-level abstractions hide these complexities, developers working in low-level languages like C++ or Rust must account for non-deterministic outcomes that defy traditional sequential consistency models.
Key insight: Even on modern hardware, it is possible for two threads to both read initial zero values from shared memory locations despite having already executed write instructions, because hardware store buffers delay visibility to other cores.
Titans: Learning to Memorize at Test Time (Paper Analysis)
Yannic Kilcher
Dec 14, 2025
Google's new Titans architecture aims to overcome transformer context limits by enabling models to 'memorize' information at test time. By using a neural network as an active memory bank, the model learns to store and retrieve past data dynamically. While technically impressive, much of the underlying logic mirrors established concepts like gradient descent and linear transformers.
Key insight: The authors frame their memory update process through the lens of 'surprise,' yet this is functionally equivalent to standard gradient descent with momentum.