Public-Diary

If I am going to turn my sporadic ramblings into something worthy of a Substack, I suspect I need to begin by being more honest. Not performatively but just honest in the way private thoughts are. In lieu of overthinking the format, I have written down a reflection on something I am currently moving through… the career transition identity crisis of one’s mid-twenties. There is a strange grief in leaving a prestigious firm. On paper, you lost nothing, in fact you gained experience, relationships, and retained or even grew, your ambition. The challenging part is loosing the answer to the question. ...

I am currently in the process of building out a messy little constellation of personal projects. Perhaps better described as instruments for thinking. They are things I want to exist because I keep reaching for them, and so likely will be worth sharing when they exist, in case others are reaching for them too :) These projects include: A database-of-databases: and eventually a path into backtesting workflows. Interesting projects like Awesome Public Datasets have previously been carried out, at great scale & with community contributions, but there is a gap I am hoping to address. ...

Science progresses via a collapsing of ontology into the new. A new framework emerges that alters how we define reality (ontology). The old ontology collapses under the weight of what it can no longer explain. The new order of meaning emerges, whereby yesterday’s observations are redeemed from their isolation and reinterpreted as signs of a deeper order. Examples of this include the shift from the geocentric (Earth-centered) to the heliocentric model of the universe. Or the transition from Newtonian mechanics to Einsteinian relativity. ...

Florilegium: literally meaning “a gathering of flowers” in Latin. In the Middle Ages, this was a book that compiled the finest, smartest extracts and quotes from various writings. Cool things I found this week… I. This personal design portfolio site that carefully documents the process of each work II. AND this personal blog website that documents their friends stories. I love everything about this. Not just the design but the approach, the realisation that our most valuable knowledge is not our own but in our friends, this person’s mindset is incredible. ...

An N-of-1 trial (or single-subject design) is a personalised clinical trial where a single patient acts as their own study population. Instead of testing a drug on a large group of people, an individual receives sequential, randomised treatments or placebos over time to determine the optimal care. N-of-1 trials are a core tool in precision medicine to evaluate customised and off-label therapies when a traditional large-scale clinical trial isn’t possible, especially in ultra-rare diseases or chronic conditions. ...

Interest in blogging over time: Blogging is still far below its historical peak, when analysing trends across a longer hoizon, but its recent rebound is statistically unusual relative to the last five-year baseline. Google Trends do not represent not raw search volume. Each point is divided by total searches for that geography/time window and scaled 0-100, so the spike means “blog” became much more prominent relative to all searches, not that we know absolute search count. That said, the spikes are interesting to investigate with context. ...

Journal Entry: The Reflexive Loop In the stillness of observation lies the foundation of understanding. Yet, what we observe is not the whole of what is. It is merely the raw clay, waiting to be shaped by the mind. Echoing the words of Heisenberg from his 1958 book Physics and Philosophy" What we observe is not nature itself, but nature exposed to our method of questioning.” Kant suggests we do not see the world as it is in itself (the noumena), but rather as it appears to us through the lens of our own cognition (the phenomena). The clay is shapeless until our minds create a form for it. ...

A system about systems of thinking. I now have a METABRAIN file in my second brain system, a catalog of every major knowledge architecture I could find: Karpathy-style AI repos [I,S,R] Zettelkasten [O,R,M] PARA [I,O] Evergreen Notes [O,M] Digital Gardens [O,S] Molecular Notes [O,S] Knowledge Graphs [R,M] Agent memory systems [R,S,M] Research pipelines [I,S] Vector retrieval architectures [R] These systems are solving different layers: Ingestion [I], Organisation [O], Retrieval [R], Synthesis [S], Memory [M]. ...

Things I found this week… I. Carl Jung on the mountains II. @kejunYing weekend project Symbol becomes structure. Meaning becomes shape A tiny weekend project :) pic.twitter.com/xQ8k7oAiFk — Kejun (Albert) Ying (@KejunYing) April 26, 2026 III. S2Vec learns city structure from map data alone, creates embeddings that predict income, population, emissions, even in unseen regions. Best results come when combined with satellite models like RS-MaMMUT, maps + imagery together. IV. Kawase Hasui woodblock prints ...

I spent a couple of weeks building a model to predict which oncology drugs would advance from Phase 2 to Phase 3. It didn’t work. Walk-forward AUC sat between 0.50 and 0.55 across every feature combination I tried: pathway position, publication sentiment, genetic evidence, novel-MOA flags, sponsor identity. None of them generalised out of sample. But there was a chart I liked too much to let die at the same time. ...