@@ -93,3 +93,6 @@ Plot Pareto frontier of journey time/cost for two endpoints.
}
-* Estimate delay likelihood based on historical data regressions and (on the day) signalling state.
+* {
+Estimate delay likelihood based on historical data regressions and (on the day) signalling state.
+* Travel for free using services which are going to be delayed.
+}
}
@@ -142,2 +145,6 @@ Automate something like https://norvidstudies.substack.com/p/juxtapose: generate
* Modernized, better [[https://www.sciencemuseumgroup.org.uk/learning/resources/launchball-game|Launchball]].
-* Like Vitalik's travel time scaling thing, use similar data to fit a frequency-of-visit-by-distance scaling law.
\ No newline at end of file
+* Like Vitalik's travel time scaling thing, use similar data to fit a frequency-of-visit-by-distance scaling law.
+* TeraCLIP: scale model (ideally) and add more modalities like [[https://github.com/facebookresearch/ImageBind|ImageBind]]. I should be able to jointly embed mathematical proofs and gamma ray spectra.
+* Scrape all recent news articles via Google News's RSS feed, download the images from them and select all the graphs from them without any context to produce a Contextless Graphs Newsfeed. This didn't work last time I tried it, I think because the news didn't have enough graphs, so get more news.
+* OutgroupRageTok.
+* Estimated reading times for blog posts based on LLM-estimated post entropy.
\ No newline at end of file