Little bit of private work. Focused on myself a lot today.
Upgraded wp to 6.2, theme (2015) to 3.4 (and removed footer).
Ordered fresh.
Indictment = getting charged. Arraignment = court to hear charges, enter plea.
GPT.
Generative Pretrained Transformer.
Many many people misunderstanding. But good that AI is getting a lot of credit.
Interesting option: https://jamesturk.github.io/scrapeghost. You basically just define a schema of data you want, and it will try to extract that from the site(s) without you writing any custom code. No DOM walkers, parsing, etc.
In general, ChatGPT should be used like google with one extra benefit: following questions with contexts. Much better at refining your query.
Another general praise: It significantly lowers the emotional resistance to creative tasks.
Switched gym task to daily (added some details), back is finally good enough to hit core/legs/yoga/other.
SBSC.
Wrote a discretization function used to bin numbers into ranges. Reducing the precision helps blend the proper groups of data. Should get better statistical analysis out of this. Used for lines (eg 3-6 instead of 5.5).
Rearranged the routes a little bit. It’s now just an SBSC page and an NFL page.
Auto-inferred a scrollbox ToC at the top of each stats page, dynamically based on plot titles below.
Started doing some neck mobility exercises. Legs and back are getting solid again.
Private work.
Trip planning. Bought LiB ticket and one-way flight (might visit Eric and/or parents and/or Eurotrip after).
A few more comms with Rise and MS.
Lots of SBSC work.
Finished all the typing, linting, and reorganizing of the manual plot builders.
Looped through the classes of statistical variables to dynamically produce the plots.
Thought about adding enums for categorical dimensions (like dayofweek). Ended up deciding it wasn’t necessary. There are two numerical arenas:
Aggregation. But aggregating function just has to be compatible with aggregated values. So mean() works with floats, but not with str values. Other aggs are fine with strings though.
Vlines. The plot shows an average for each trace (of the main dimension). But these are always numerical. Sure, our main dimension can be categorical (cover/push/noncover), but the values will either be counts or percentages.
Wrote a function that (recursively) converts (nested) defaultdicts to dicts.
Everything ready for optimization now. Hopefully will finish tomorrow, which would conclude this 3-week ticket.
The muscle in my ear that I can vibrate is called the tensor tympani (TTM).
Private work.
Github copilot (powered by gpt3) vscode extension requires a github account with access to copilot (obv). It’s $10/mo.
Remember I currently use tabnine.
Was able to have a few conversations and feel a little normal a couple moments today.
California Department of Financial Protection and Innovation (DFPI): https://dfpi.ca.gov/
Nationwide Multistate (redundant) Licensing System & Registry (NMLS): https://mortgage.nationwidelicensingsystem.org/Pages/default.aspx
Some good progress on SBSC. Basically finished the dimension definitions, the generic builder, and the generic plotter. Hand wrote the argsets. Now just need to classify the dims then dynamically loop (then final optimization).
Connected with Rise again, dropoff for good morning america will be in 2 weeks.
SBSC.
Deleted the constants module and moved all vars to their appropriate modules (everything was single-use).
Moved COLOR_TRIO to the dimension definition where it belongs, not a default assumption of plotting utilities.
Removed sort_by. Everything is by trace value, sorted in trace order (first name = primary sort, second name = secondary sort, third name = tertiary sort, etc).
Made the xaxis (main dim) able to handle many many many traces:
This shows every line and the % of every team’s games that were that line (you could plot counts instead of %). The actual (valuable) plot uses the aggregation utility, of course – we care about the avg line by team.
