I am on a tightrope with my wife. Balancing between “this is too weird” and “this very helpful”.

Mason Veilleux@mveilleux.com · yesterday

I made a family chat bot for my wife and I (using Hermes) and by far the most fun part has been developing the voice and lore of the bot. I have the bot say good morning and recount the dream it had, example:

I made a family chat bot for my wife and I (using Hermes) and by far the most fun part has been developing the voice and lore of the bot. I have the bot say good morning and recount the dream it had, example:

Recently found out someone I know is deeply involved in that hippie enclave in Oregon from the Wild Wild West documentary. And yes present tense because the community is still very much alive.

If you wanted to know Windows Server comes also with Xbox programs and this is very critical to enterprise operations

Mason Veilleux@mveilleux.com · last wk.

Honestly really love how Adobe completely hijacked my screenshot tool. Oh and also the Lenovo ads that overlay where my screenshot appears.

Honestly really love how Adobe completely hijacked my screenshot tool. Oh and also the Lenovo ads that overlay where my screenshot appears.

You can just commit your .env files

Once you understand the Minions are Italian it all makes sense

Some friends to moved to Germany (German citizens) so why not.

FG model is looking money. Here is kicker skill distributions over time and centered.

I’m going to be insufferable in a few Sundays

Cool ending to halftime show but those kids were pitchy af

The US Government doesn’t want you to know that C-pop slaps

Sports data science jobs are either taking a considerable pay cut to work for a reputable sports team or sell your soul to sports gambling to run experiments on SEC frat boys

Hear me out. A website of a Bradley-Terry model for every sport.

1. I can write software 2. I can make eye contact to who I am talking to I am proud to announce I am a forward deployed engineer

Also, PyMC has been incredible. I can tell it to run all of my sampling on GPU (12GB) (goes brrr with larger geometry), then I can set a flag to do all the post-processing on CPU (now using RAM with 64GB) with : postprocessing_backend = “cpu”

Learned something new in Stan where you can have two priors — one for regularizing and the other for the actual data you want to inform the model with.

Couple of dad L’s at the park. Head up high. Ball up top.

If you know enough Elixir they just give you a Brazilian passport

Thank you PyMC

How many of these can I get before it gets weird?

So I guess “they” are gatekeeping blood boys now.

Ally Whelan@amwhelan.bsky.social · 4w ago

Excuse me, what? This needed to be studied?

Does being senior mean I get my own blood boy, or do I have to share till I reach staff?

Un-un-pivoted. Double Reverse Uno.

Mason Veilleux@mveilleux.com · last mo.

I am not pivoting to life insurance

The EOD job applications hit the hardest.

I am not pivoting to life insurance

Mason Veilleux@mveilleux.com · 2mo ago

Might pivot to selling life insurance

Stan is dead. It’s never been more over.

Mason Veilleux@mveilleux.com · last mo.

Pymc with backend JAX on GPU goooo brrrrrrrrrrr

Pymc with backend JAX on GPU goooo brrrrrrrrrrr

All this to be able to fit this better:

Mason Veilleux@mveilleux.com · last mo.

I ended up finding a way to do a partial proportional ordered logistic model. Basically needed to do this where each yardline category needed to have a different effect on each drive outcome (yardline effect is different for a TD then for a safety).. didn’t even know that was allowed.