Nourdine Lotfi

harn. full Python port of earendil's pi agent

rubrics. a bunch of rubrics I made in different format and structure for llm judge and other use cases

niwa. niwa(庭) — Async conflict-aware spec/planning, for me, you and your LLM/agent

rubrify. Rubric compiler and judge engine for LLM evaluation

claude-code-ui-freeze-investigation. Investigation and fix for Claude Code v2.1.23 UI freeze bug

goal. /goal port from codex to arbitrary agents and harnesses, opencode, claude code, etc

scraping_work. Python

unsloth_multigpu. Finetune Llama 3.3, DeepSeek-R1 & Reasoning LLMs 2x faster with 70% less memory! 🦥

elwood. Programmatically import and instrument the locally installed Claude Code CLI via Babel AST -- drop-in replacement for @anthropic-ai/claude-agent-sdk

audiograbspace. Chrome extension to download x.com audio space recording

arxiv-tab-downloader. Download all opened arXiv urls

git-autoeverything. Git autocommit & autosync script

obsidian_excalidraw_tools. just some tools/scripts for my workflow with obsidian+excalidraw

jari. jari / 砂利 — Task/issue tracker for LLM agent workflows — LMDB-backed, priorities, dependencies, conflict detection, niwa integration

tkinter-repl. Making different kinds of REPL using just tkinter and local python libs

screen_rec. Screen Recorder supporting various libraries

manifold-transformer.

sudoku_solvers. Python

Stuff2C. Random stuff I needed to be in C for curiosity (and speed)

rubrify. Python toolchain for XML-first LLM rubrics: generate, evaluate, calibrate, constrain, and evolve prompt programs.