A collection of technical write-ups, engineering notes, university experiences, and thoughts.
What happened when I fixed reflection blind spots, resolved sandbox import bugs, and scaled Patchwork from 3B (60%) to 7B (76%): which bugs were agent flaws, which were model limits, and how confirmation bias shrank.
How an execution-guided LangGraph reflection loop boosted a 4-bit Qwen2.5-Coder-3B model from 38% Pass@1 to 58% Pass@overall, and the confirmation-bias blind spot it uncovered.
Engineering an asynchronous LangGraph pipeline with Groq 20B/120B dynamic routing, Qdrant vector retrieval, and thread-offloaded cross-encoder reranking.
Why Python's str.rstrip('.com') does not remove a suffix, how it silently mangles strings, and why 3B-parameter models hit 0% accuracy on string stripping.