f311a

7 ms·

61.
by f311a 3mo ago
The real price is how often a model uses subagents that scan your whole repository multiplied by thinking level.
62.
by f311a 3mo ago
I'm working on hexora, a library that detects malicious Python code using static analysis and machine learning. Recently, I've added a simple ML model to filter out false positives. In the last month, I found and flagged more than
63.
by f311a 3mo ago
It's not idiomatic implementation in Rust, just translation from C. A few Rust macros simplify implementation and cut the code in half.
64.
by f311a 3mo ago
The code is so weird, people will have a hard time reading it when they need to check for correctness. There are much better ways to write it in Rust: https://github.com/malisper/pgrust/blob/14ffab7d31a31e5ab6
65.
by f311a 3mo ago
Yeah, claims that it can be faster than CH are very suspicious. CH guys are very good at their craft, they spend hundreds of hours optimizing one single small detail.
66.
by f311a 3mo ago
I think Google stopped using such data long time ago. They have been showing hydrants, buses, crosswalks for ages. It's a solved problem to detect them.
67.
by f311a 3mo ago
They don't have a goal of becoming a popular language, though and will continue to work on it as long as there are donations. They don't care about being mainstream and there are niche companies who appreciate Zig and donate.
68.
by f311a 4mo ago
x4 pro will come with backlight and touch screen https://www.youtube.com/watch?v=REKdjxYtMrI
69.
by f311a 4mo ago
It's the same on any package index now.
70.
by f311a 4mo ago
> What does any of that mean in practice? They want you to spend more tokens
71.
by f311a 4mo ago
It’s a slop article based on a tweet.
72.
by f311a 4mo ago
No, it's good for coding and faster to iterate. Flash models work almost the same for changes under 1000 LOC when you dictate how you want your code to look. Big models still suck at big changes and big changes are hard to review.
73.
by f311a 4mo ago
I'm using Deepseek flash for certain tasks, it is pretty hard to spend $10 on it. All their top models (by usage) are not from frontier labs.
74.
by f311a 4mo ago
Exactly, I also prefer when neighboring cells are automatically marked when you place a star
75.
by f311a 4mo ago
I reported 100 of them 4 months ago and they were not removed after 3 months. They just don’t care, no one answered the ticket.
76.
by f311a 4mo ago
I wish this article was not LLM written
77.
by f311a 4mo ago
It's used in ClickHouse as well. CH supports all known compression algos and they are documented pretty well.
78.
by f311a 4mo ago
> If you can find resources to learn something that was written by a human, read that first. But Lathe is here to fill in the gaps when that isn't the case Well, but it will still serve you content from humans, but without any attri
79.
by f311a 4mo ago
You don’t need to connect to duckdb, it’s just a process that you spawn.
80.
by f311a 4mo ago
ClickHouse also supports a lot of data sources and has a local mode where you just use a single binary with local-only access. Coincidentally, I wrote an article today on how I use it for similar scenarios. It can fetch from S3, multiple da
81.
by f311a 4mo ago
It used to allow reading replies without being signed in. Not sure what changed, but now it just redirects me to x.com.
82.
by f311a 4mo ago
It runs in a terminal, it’s not electron
83.
by f311a 4mo ago
Infrastructure is a much harder problem. They can't even improve Claude Code, which eats 1GB+ of RAM. Meanwhile, my editor only consumes 80MB of RAM.
84.
by f311a 4mo ago
I hope he does not use it and just wanted to advertise his project to get some Github stars...
85.
Ask HN: How do you find deep technical content?
40 points by f311a 4mo ago|28 comments
86.
by f311a 4mo ago
You can ask LLMs about high-level techniques, and their answers will usually be good enough. What you can't get from LLMs is the taste and judgment, which you can only obtain by having a strong CS base and coding manually for years. Hi
87.
by f311a 4mo ago
Can you show me an example of a hard task that can't be achieved using light models? When we don't want the model to work on autopilot without reviewing the code at all. Even SOTA models will produce garbage code, if you don'
88.
by f311a 4mo ago
How many more months do we need to wait, until big companies realize that flash models work just fine if you: 1) Don't ask LLMs for big changes 2) Review everything and point them in the right direction Large models still suck at big c
89.
by f311a 4mo ago
This web site is very hard to read because of the colors and font sizes.
90.
Cessation of public development of Kefir C compiler(kefir.protopopov.lv)
141 points by f311a 4mo ago|134 comments

More ›