If you see a gap month or year between these, it’s either I am legally bound to keep it shut, or a personal project or it’s just being human and staring at the wall or resting. You know, the latter somehow always satisfies my criteria for choosing problems perfectly.

June 2025:

I was working on a frontend design for the first time. The user was someone who just want to get off work and go home, a tired user on friday 6 pm. But usual visual standards are 16px font for base text, the color pairs being actually readable hanging off by dreams and prayers and hoping it works (tho it looks really great when coding it and looks aesthetic and COOL)

I wondered “why not let the user have the option to tap a button and let it change to a higher contrast, font” and just easier for the user to go through the app.

That became my first research work:

Beyond Compliance: A User-Autonomy Framework for Inclusive and Customizable Web Accessibility

Well well, I found out a month or so later about the whole horrible experiences people had with something called an “overlay”.^{1 2}

Basically that was the idea I converged to as well. So the implementation part of the paper is to be considered BS, but the underlying principle: “The complaince to make the web usable to people by what law mandates is the floor” still stands and is something I believe to this day. ³

How did this end? See Decemeber 2025.

August 2026

Talked to few folks in legal tech space. Considered a problem at a legal tech firm, and there was a cool overlap with the work, they have been converging on law as groundtruth and analytics, something close to one of my previous works. It produced A founder’s guide to building on public court data without walking into a wall (more on that below).

But the problem fit someone who’s more of going from 7-9 (maintaince, uptime and reliability) than 0-1, I am more interested in the upstream of this where we define and calibrate what should be maintained, and what risks are involved, etc. So it was a no.

Now onto that guide, it was more of a quick of risk analysis I did before meeting, but here you go. This is written from the POV of building using court data — a few things worth knowing going in:

• Public data isn’t automatically free to build on commercially. What matters is what you do with it, not just where it came from.
• Naming a lawyer or litigant is generally fine. Scoring or ranking them without their input is where actual lawsuits have happened elsewhere.
• Judge-level analytics — even indirect, inferred patterns — is the one thing that’s already led to an outright ban in at least one country.
• If you ever partner directly with a court on an AI system, current draft rules suggest the court can end up with a perpetual, royalty-free license over what you built — worth knowing before that conversation starts.
• The same underlying data and analysis can be low-risk or high-risk depending entirely on how it’s packaged and shipped, not on the data itself.

Came across an interesting role at an LMS platform I have been following with thousands of students, there were really interesting rabbit holes (do you make an exam scrollable or one question at a time?) and the team were wonderful. But the gap being filled was more of execution than where my work would be a good fit.

September 2026

Considered a mentor/teaching role engagement for one of my fav subjects, but it did not align on one of the heuristics- #2, so it wasn’t a good fit.

This page is a work in progress (gotta print context out of the brain). Expected time for it to be updated is: Unknown and depends on the universe.