I'm considering polars as my next "special interest".
Step 1: use it daily; play
Step 2: pull the hood up and figure out the codebase
I've seen hints of cool DB and engineering concepts from reading the codebase, but i imagine there's a lot i can learn here if i STUDY
@GaelVaroquaux I had a very young and nervous econometrics teacher who struggled so much with “heteroskedasticity” and “multicollinearity”: the whole class was very quietly rooting for him each time
Had a very interesting bug where ghostty couldn't `ls ~/Downloads` but <other terminal> could
Reminder to self to dig into "what is happening? why?"
I feel like I'm getting a lot more rate limiting from Claude today...I think I'm going to start moving a lot more of my dev to open weight models and simple harnesses if I can
Markdown vs HTML. Every time we go from a semantically dense to a semantically sparse format, we lose. Even more today where less tokens from the same content is way better. I can understand we need a better markdown. I can't understand we should replace it with HTML.
@MelkeyDev I do my best to stay under 100k: limit context “manually”, start a forked session if I entertain multiple ideas, or frequently invoke subagents. It becomes easier to have tightly scoped sessions if I use something like https://t.co/R4yjfeHD8q for commiting frequently
new release:
NEUROSIS "An Undying Love For A Burning World"
新メンバーとしてAaron Turner加入、NEUROSISが衝撃の再生を果たす���
完全新録アルバムをサプライズでリリース。
日本盤CDを含むフィジカル・リリースは5月中旬以降予定
https://t.co/uq1p7K9H2G
@shadcn lol though that actually looks like i didn't understand it yet (and probably still don't)
(1) it should be `git worktree remove .worktree/hotfix`
(2) you can't check out multiple versions of `main`
I wonder why I don’t see much about SSMs for CLV; a lot of the Bayesian marketing models have distribution assumptions that seem nice on paper but break really easily
just got my first PR review by a human that was clearly all done by an agent ....very weird feeling
some comments are useful, most are crossing "T"s and dotting "I"s, but the biggest impact is eroded trust in the reviewers' judgement :/