I'm 70. I've been building a set of tools and websites that didn't exist and that I thought should. What follows is an honest account of what's actually on my workbench right now, written for someone who's curious and intelligent but doesn't need the technical details to get the point.
The web rewards the already-curious. If you search for what you already know exists, you find it. What you almost never find is the thing you didn't know to look for — the paper, the obscure blog, the dataset, the writer — that would have changed how you think about something you care about.
I'm building a system I'm calling the Wanderer. It monitors a curated list of sources — academic preprint servers, RSS feeds from specific thinkers, Wikipedia's new articles, GitHub trending, a handful of Substacks — and scores each find against my personal reading history. Things I would predictably already know score low. Genuine surprises score high.
The historical anchor for this idea is the flâneur — Baudelaire's term for the urban wanderer who follows curiosity without agenda, collecting impressions, noticing what others stride past. Alexander von Humboldt spent five years crossing South America doing essentially this at a continental scale. The connections he found — between altitude and temperature and plant species — nobody would have found by staying in a library. The surprise is the point.
I work with AI assistants every day — primarily Claude. The problem I've run into is one nobody talks about cleanly: I have far more projects than any single conversation can hold. An AI assistant that knows about my book catalog doesn't know about my bioengineering page. One that knows the network topology of my home lab doesn't know what I decided last week about the civic information site I'm building.
The standard answer — "just remind it each session" — doesn't scale. It's like having a brilliant collaborator who loses their memory every night and makes you re-introduce yourself every morning.
I'm building a different approach: a topic registry, stored in a database, that maps everything I'm working on to named topics. Each topic has a status, a heat score, and a summary. At the start of each working session, only the active, relevant topics load into context. Stale or dormant topics stay in the database, available when needed, out of the way when not. The goal is a working context that stays current without requiring me to manage it manually — the system handles the swap.
It runs on Cloudflare's infrastructure, which means it costs almost nothing and is available everywhere. The session handoff protocol is already working. The topic registry is the next piece.
I've built nine websites. Each one addresses a question I couldn't find a good answer to elsewhere. Taken together, they're an attempt to build something I'd call civic information infrastructure — tools for thinking clearly about the world in a period when that's becoming harder, not easier.
The sites are listed below. Each is a different facet of the same underlying concern: what does a well-informed, intellectually honest person actually need to think well? Books, yes. Historical context, yes. An understanding of how democratic institutions break down, and what the founding generation thought about that. Systems thinking. Mathematics as a way of seeing. A catalog of ideas, not just facts.
Each one is a standalone resource, but they're designed to link to each other. The cross-links are intentional — the same idea showing up in a book, a political pathology, and a mathematical structure is the point.