HackerTrans
TopNewTrendsCommentsPastAskShowJobs

jt-hill

no profile record

Submissions

Thoughts on OpenAI acquiring Astral and uv/ruff/ty

simonwillison.net
86 points·by jt-hill·قبل 4 أشهر·24 comments

You don't hate Python. You hate other people's Python.

jt-hill.com
8 points·by jt-hill·قبل 4 أشهر·6 comments

How does AI detection work?

pangram.com
1 points·by jt-hill·قبل 4 أشهر·1 comments

Decimal Is Sometimes Wrong

jt-hill.com
10 points·by jt-hill·قبل 8 أشهر·3 comments

Show HN: Credkit – Easier Loan Modeling

github.com
1 points·by jt-hill·قبل 9 أشهر·1 comments

comments

jt-hill
·قبل 29 يومًا·discuss
Always read the PKGBUILD. Always read the updates diffs. It’s that easy.
jt-hill
·قبل شهرين·discuss
> I have no idea if it works or not because I haven’t yet tried running it.

This should have been disclosed in the first paragraph
jt-hill
·قبل 3 أشهر·discuss
Wow it just keeps going. Extremely thorough.
jt-hill
·قبل 4 أشهر·discuss
This is a relief in many ways. With Fivetran’s acquisitions of both dbt labs and Tobiko, I was expecting the end of the open source data ecosystem.
jt-hill
·قبل 4 أشهر·discuss
Astral was always going to have to find some way to sustain itself financially. They weren’t going to just make the best free tools in the ecosystem forever. uv is sufficiently entrenched as infrastructure that I’m sure it’ll take no time for a community fork to show up if they do anything stupid with it.
jt-hill
·قبل 4 أشهر·discuss
Ruff solves this. You should not be manually fiddling with indentations in 2026.
jt-hill
·قبل 4 أشهر·discuss
Hah, thanks. Believe it or not I proofread it like three times but my mental autocorrect is just really aggressive.
jt-hill
·قبل 4 أشهر·discuss
I keep making this argument elsewhere on the Internet, so I thought it would be nice to make my point all in one place. Python is good now, if you weren't paying attention.
jt-hill
·قبل 4 أشهر·discuss
Related recent discussion: https://news.ycombinator.com/item?id=47202864
jt-hill
·قبل 8 أشهر·discuss
Thanks! And you're right this is true for storage and at the transaction level where nothing smaller than, e.g., a penny is allowed to move or exist in the first place. But when you're working with interest rates and the main calculations involve multiplying by numbers < 1, you still have to figure out rounding.
jt-hill
·قبل 8 أشهر·discuss
Author here. I thought was doing what I was supposed to do by using Decimals heavily in my credkit[0] Python library, but I did some testing and discovered it wasn't worth the pain for me. Test code [1] is available for the curious.

The moral of the story is that "best practices" are not a substitute for using your brain.

[0]: https://github.com/jt-hill/credkit

[1]: https://gist.github.com/jt-hill/839e1a6c51d28cbcb8fc69e61431...
jt-hill
·قبل 9 أشهر·discuss
I work in credit risk and needed to model consumer loans in Python. I couldn't find a tool that made it easy to model simple loans either individually or at the pool level, so I decided to make my own.

It's early (v0.2.0) but works end-to-end and can produce an amortization schedule from loan characteristics

Up next are the actual credit risk features I wanted from the start.

Feedback welcome!