Timeline
What actually happened. Talks, events, things shipped, books that mattered, and what I was learning at the time.
A record rather than a plan. List 99 is what I want to do and five years is where I am going. This is what has actually happened, which is a different and more honest thing.
Books appear here when I finish one and it changed how I think. The note says why, not what it was about.
15 entries 3 events · 5 milestones · 2 books · 4 learnings · 1 writing ·
-
learning
Robin Linacre on probabilistic record linkage
The clearest account of the Fellegi-Sunter model anywhere, and free. Working through it properly rather than by osmosis, because I have been using the vocabulary for longer than I have understood the maths.
-
learning
Apache Ossie, the semantic interchange spec and semantic layer
Formerly Open Semantic Interchange, donated to Apache and incubating since June. A common format for what a metric actually means, so the same KPI does not quietly mean three things in three tools. Close to my day job and close to the knowledge graph question.
-
learning
Correlation clustering, and why transitive closure ruins a graph
Similarity is not transitive, so pairwise scores cannot produce a partition. Still working out what to do about it.
-
milestone
Joined the PKI Consortium
To learn public key infrastructure properly rather than by osmosis. Narrow interest, non-repudiation and confidentiality.
-
milestone
Started the five years
Picked a direction and a clock. The number in the footer counts from here.
-
book
The Infinity Machine — Sebastian Mallaby
Read it for Demis Hassabis and what it takes to point a research lab at one very long problem. The interesting part is patience, not intelligence.
-
event
ID4Africa, Abidjan
The Parc des Expositions, 12 to 15 May. The continent's identity people in one room. Useful mostly for what it made obvious: everyone is building registries and almost nobody is talking about what happens when two of them disagree about the same person.
-
book
This Is For Everyone — Tim Berners-Lee
His own account of the web. The line that stayed with me is that the architecture was a choice, which means the current shape of it is also a choice and could be a different one.
-
milestone
Data for the Rest of Us published
Troubador. Around fifty copies out so far, which is fifty more than none.
-
writing
Identity is an inference problem
The post where the thing I actually care about finally had a name.
-
event
Connected Data London
Graph people, and the first time I heard the identity problem described in someone else's vocabulary. Good for realising a thing you thought was niche has a whole field attached to it.
-
event
AfroTech, United States
Different room entirely. A reminder that the technical problem and the question of who gets to work on it are not separate problems.
-
milestone
Published graphfaker
A Python package. Second one out in the open.
-
milestone
Published overturemapsr
An R package for Overture Maps data. The first thing I shipped that strangers used.
-
learning
MSc Information Management, NOVA Lisbon
Two years. Where the vocabulary for all of this came from.