Give Your AI a Watch
Why AI agents need persisted message time outside the system prompt.
Posts
Engineering stories, experiments, and working practices.
binary models 1
CUDA 1
llama.cpp 1
local models 1
memory 1
quantization 1
research 1
software engineering 2
vs code 2
vulkan 1
working practices 2
zed 2
AI Mid-Memory: From a Goldfish to a Personal Assistant Worth Naming
A practical system for recovering AI sessions, maintaining reviewed memory, reusing concrete workflows, preserving provenance, and evolving an operating contract.
LLM Binarization — Episode One: DBF aka Addition Is (Almost) All You Need
A runnable DBF/SIGN1 model: near-Q3_K quality at a slightly smaller-than-Q2_K budget and near-Q2_K Vulkan speed.