A daily, dated record of what I'm reading and building. One entry per day, committed automatically each morning (IST).
Each entry highlights up to 3 recent papers on arXiv in cs.LG (machine learning) with their titles, authors, links, and short abstract excerpts — a genuine, verifiable learning log rather than padded filler.
devlog.md— the running log (newest entries appended at the bottom)daily_devlog.py— the script that appends today's entry and commits/pushes itbacklog.json— pre-staged entries for upcoming days (local only, git-ignored)
python3 daily_devlog.py --backlog N pre-stages real arXiv entries for the
next N days into backlog.json. On each day's cron run, if an entry is staged
for today it is consumed and committed without any network access — so the
green square survives WSL being off or arXiv rate-limiting during the fetch
window. Each day gets its own slice of the feed so papers don't repeat.
Consistency is the point. A green contribution streak built from real, dated entries signals steady, disciplined work over time.