Founder, Civic Literacy Initiative. Former data lead @okfn, @schoolofdata. Making sense of the collision btwn tech & society through (data/algo) literacies.
@mattbeane I stopped at 10' when it said "5 genres, 5 more to go" before suddenly skipping to the next paper. I would have expected more callbacks to the first paper being discussed while discussing the second one. It felt like I had to do the connections myself, which I find distracting.
Text is a summary from a section of a series of lectures (in French) by Alain Supiot, titled "Governance by numbers" (7th lecture). A companion book was edited, and it received an English translation https://t.co/wUsunYTawQ
Highlights from https://t.co/MmKAGs9FzY
The rise of social stats led to two interpretations
— those using them to identify social ailments and advocate for change
— those twisting them into flawed biological conclusions in order to advance practices like eugenics.
(Alain Supiot)
The more things change...
@evidence_dev Would be great to have access to an RSS feed for your blog! Much nicer to follow product updates from my RSS reader than across email / social media.
One of the most inspiring sentences I've read recently.
Also thankful that this is much more than a lucky strike of brilliance: @adriennemaree has developed the topic across several publications—all promptly added to my reading list.
@archieemwood@evidence_dev What I mean is that the philosophy behind Evidence seems to lean toward consuming the db and writing the SQL in Evidence, rather than generating a Datasette endpoint that I consume in Evidence. Or is it?
@archieemwood@evidence_dev But reading how evidence works, is there any point in hooking it up to the API if I have access to the SQLite behind it? As Evidence stores stored and converts all the data anyway?
@archieemwood@evidence_dev I've been using Streamlit to create small internal apps that consume a Datasette API. And I've been meaning to test Evidence because of how slow Streamlit can get while caching the data.
@archieemwood@evidence_dev The user here is the biz analyst building the data-driven website right?
In this case an interesting use of llm would be to generate x different type of charts based on the same data to compare readability.
UI could be freeform or use dataviz grammar to help guide prompt.
@ethanf_17 Is there a reason to extract and transform the data directly with DuckDB instead of doing it with pandas (or polars) and then loading it in a DuckDB file?
@simonw That allows them to deploy quickly to demonstrate the use case, then find funding after. Sounds like a good use case for serverless to me? Though not sustainable for the platforms themselves probably.
@simonw I had a meeting yesterday with a gig department that was responsible for updating a CSV data file but had not control over its publication, as another department managed the open data portal. I suggested Datasette + Vercel to them as a way to circumvent the issue.