Jim Clifford
@jimclifford
Professor of History, University of Saskatchewan Co-Editor of Historical Methods. Environmental and digital historian of London, Canada and the British World in the 19th century. #dyslexic
Using data from a published paper, I've visualized settler arrivals at the Port of Quebec 1817-1839. The website includes a map of trade networks and a case study of Westmeath, where some of the people settled in Algonquin territory. jimclifford.ca/quebec/
I just finished The Soul of the New Machine by Tracy Kidder. It is wild how similar the concerns were 45 years ago. AI, job loss, increased control for management, power of tech leaders, and just about every other concerns we have today.
Sharing one more project that Fable 5 helped me complete today. The Timber Trades Journal recorded the arrival of every timber ship in every port in the UK and we OCR'd the pages and parsed the data into a knowledge graph and visualization. #envhist #digitalhistory jimclifford.ca/timber-data/
I'm also working on a map of the locations that include latitude and longitude in the text. jimclifford.ca/britannica-w...
There is a visualization to show the changes in each article between the editions.
Working with the National Library of Scotland data, I've reOCR'd all of the Encyclopædia Britannica 1771–1860 and parsed the articles so we can see the change in a topic's coverage over 89 years. I've been working on this for months and the Fable 5 model solved a few remaining major issues.
There is a lot to this question. The Americans figured out timber framing relatively recently (Kate 19th century). Before material science developed, everyone over built structures. The British imported Baltic fir dating back to the late Middle Ages. The supply extends to Canada, Russia etc.
An interactive atlas of the British Empire's civil service: ~16,000 Colonial & India Office careers (1820–1966), every move an arc across the map. Rebuilt from the digitized Indian and Colonial Office Lists with a knowledge graph. WIP, feedback welcome: jimclifford.ca/col_matching/
It took me months to figure out how to extract 30,000 careers from the Colonial Office List; it took a little over a day to repeat the full process for the Indian Office List. The visualization now shows the very different global geography of these two bureaucracies in the British Empire.
After creating a visualization of mobility between Canada and the empire, I decided to visualize all of the career mobility recorded in the biographies in the Colonial Office List. (I need to do the same work with the Indian Office List to complete the network.)
One question I started with when I found the Colonial Office list was how integrated Canada was in the British World System when it came to career colonial officals. #digitalhistory
I've been working with the Colonial Office Lists (1867-1965) for months. I've finally parsed the biographies of about 30,000 colonial officals into a knowledge graph. Here is the offical who worked in the most colonies during his career. Locations, institutions, and jobs grounded to Wikidata.
I'm nearing the halfway point of this impressive book. But I'm also noticing some suprising ommissions in the citaitons. Beckert has a few pages on guano and doesn't cite @guanoguy.bsky.social. He has a section of Indiginous despostion but doesn't cite Glen Coulthard (and other Indiginous scholars).
I wrote a short essay for the @nichecanada.bsky.social series on Canada and the Great Acceleration. niche-canada.org/2026/06/16/w...
Qwen 3.6 27b is an excellent open weight model. I was testing it at correctly searching and selecting Wikidata IDs to match people in the Dictionary of Canadian Biography. It only made 9 errors and on inspection, 3 were errors in the test data, reducing the count to 6. This matched Gemini Flash.
ESRI killed their original StoryMap platform and one of my earliest digital history interactive maps in the process. I managed to rebuilt it with some help from Claude Code. jimclifford.ca/london_indus...
Got it done. Boston is a hard course. It was unlike anything else I’ve done.
I'm reading the new book The Infinity Machine and Demis Hassabis and DeepMind. I didn't realize how important Canada was in the AI breakthrough, but students needed to leave universities and Canada to bring the Alberta school (reinforcement learning) and the Toronto school (deep learning) together.
Finally, here is Claude Opus 4.6 side by side with our smaller version of Qwen 3 (14b). A core goal is a model that corrects the spelling and basic grammar issues without erasing the kids voice and making it read like a generic adult. Claude and our model does well by not being too agressive.
I've been working with my son for past few months to develop an LLM that corrects kid's dyslexic writing without providing an "easy button". After trial and error, we now have a model that matches the performance of Claude Opus 4.6 on a test story. The first two images show the story and corrections
We need to change this conversation and the underlying publish or parish culture it reflects. We don’t need more papers. Could we instead aim for better papers? Claude is very powerful, but what is the point of pumping out 50 paper that the ‘author’ half comprehends?
“No one is going anywhere. Canada will survive the brief moment of honesty.”