← Ana Sayfa
// WRITING
GÜNCEL
Building agent memory that actually persists 2026.06.12 language models forget the moment they finish replying. memory is everything the system arranges around them. Retrieval beyond cosine similarity 2026.05.03 similarity search is the easy part — reranking, filtering and hybrid retrieval are where answers get good. Market data pipelines with fastapi + vector dbs 2026.04.18 streaming tick data into embeddings, and what breaks when you try to query it in real time. Evaluating llms on quantitative tasks 2026.03.02 benchmarks lie. building small, honest evals for math-heavy and finance-heavy workloads. What i'm currently breaking and rebuilding 2026.02.10 notes from the workbench — pydantic ai pipelines, agent loops, and other things held together with tape.