Two Years of Commits, Read Back
I sorted my repositories by creation date and read them in order. It is a more honest document than anything I would write about myself deliberately.
2024 — the description that aged well
C, May 2024. Repository description, unedited: "beginner programmer; want to make a working integral calculator."
I still have not finished the integral calculator. I have pushed to that repo as recently as this year.
Two months later: math-cloud, and an Expo app called crash-course-in-mathematics-for-p6. The pattern was already there and I could not see it — everything I built was maths, wearing whatever technology I had just discovered.
2025 — breadth, mostly accidental
August: a water-saving RPG in GDScript, because a friend mentioned Godot. learn_DL, whose description is just "& all that jazz" — that one is me starting Learning Deep Learning and being slightly embarrassed about announcing it.
Also August: hackit_blingbangbangborn, which became NeuPoster. My first project touching full-stack, generative models, and an ESP32 in the same repo. It half-worked.
October: dsag, algos, DSAG_C-. Data structures and algorithms in C#, which is what school looked like that semester.
December: a live dashboard of the probability distribution of discrete-choice model data, for a Google AI catalyst hackathon. First time I built something where the interesting part was the statistics and not the interface.
2026 — the year it compounded
February and March came fast. pre-hackomania-2026, then jwj2026 — which is SENTINEL — then a browser extension, then ContextGuard, then ev3 and the line-tracing workshop material.
March: hardware_projects, which is Lifeline. My first repo with no code in the usual sense. Just OpenSCAD, renders, STLs, and a technical report.
April: ProjID3, the CO₂ adsorption research, and soar-main, which is SOAR. Also landing-pages-agency and a job-hunting agent skill, both of which are me finding out what an agent can be trusted to do unattended. (Answer: less than I first assumed.)
May: MathScribe, and this website.
June: Agnes for a merchant co-pilot hackathon, SenseMake for turning student reflections into a reviewable library of values and skills, and a couple of retrieval experiments.
July: a financial literacy app for fresh graduates, Singapore Informatics League write-ups in TeX, and an R project asking what the causal effect of AI-tool use is on hours spent on schoolwork. That last one is me pointing the method at myself, which feels fair.
What the list actually shows
Three things I would not have said out loud before reading it back.
I have exactly one interest. Mathematics, applied to whatever is in front of me. Games, sorbents, LaTeX OCR, discrete choice models. The domain keeps changing and the itch does not.
I finish about half of what I start, and the unfinished half is not wasted. The integral calculator is unfinished. The ESP32 never got its bitmap. Both taught me the thing the finished version would have taught me, slightly later and more cheaply.
The gap between "watched a workshop" and "shipped a container" got much shorter. In 2025 that gap was months. In 2026 it was a weekend. Not because I got smarter — because I stopped treating unfamiliar infrastructure as a reason to wait.
The commit log is not a highlight reel. It is a record of what I was curious about on a specific Tuesday, and there is a version of me that would find that embarrassing. I am trying to be the version that finds it useful.