Why keep an open research notebook?
A paper tells the final story. This notebook keeps the questions, detours, failed attempts, and small discoveries that led there.
LucidGPT: a closer look at the training loop
GPT-style pre-training, optimizer modes, benchmark evaluation and training monitors, with an interactive workflow and a synthetic loss-curve example.
What-s-up: building and checking a chat application
A project showcase of chat software contributions in backend communication, functional testing and debugging.
AI4Game: starting a project notebook
A place for the research questions, development decisions, experiments, and lessons behind AI4Game.
How to write up an experiment that didn't work
A reusable structure for negative results: preserve the hypothesis, the evidence, and what you would try next.
From a question to a reproducible experiment
A practical checklist for connecting a research idea to a baseline, an evaluation, and a run someone else can inspect.