I keep the useful details here: why I started, what I tried, what exists now, and where the idea stops. 3 projects, in different states.
01 / Projects
These are experiments I spent enough time with to put online. They are not all finished, and that is part of the point.
Developer tooling · Working prototype
Why I thought of it: Coding agents can sound certain even when a change is broken. I wanted to see what a small, independent checking loop would look like.
What exists: A local Python project that runs declared checks, records failures, and gives the agent a bounded chance to try again.
Read the notes →Agent infrastructure · Experimental
Why I thought of it: I was curious about what happens when an AI system has to act under uncertainty instead of being told to simply be careful.
What exists: An experimental protocol with a planner, a few kinds of verdicts, and a gate that refuses to treat unknown as safe.
Read the notes →Bioinformatics · In progress
Why I thought of it: A sequencing pipeline can finish without errors and still leave behind results worth questioning. That felt like a useful place to connect my degree with my interest in AI systems.
What exists: A Python tool for reading sequencing artefacts and logs, applying explicit checks, and pointing back to the evidence.
Read the notes →More unfinished experiments are on GitHub.