f311a

Substack gives a really nice exposure though, unlike medium and other platforms.

It's very hard to gain readers now on independent websites. Google search is dead (very few people now search for tech content), social media platforms penalize you for posting links and so on.

You can get a few thousand subscribers on Substack just by being consistent and interacting with other people. Now try gaining 1000 subscribers to an e-mail newsletter or RSS from your personal blog. People who tried it know that 5,000 views on your blog can convert to a few subscribers, and you need to get those 5000 somehow first.

A lot of people just repost their articles to substack for a reason. It's not forbidden. You can repost them from your personal blog.

> less training competition

I think you meant less research and experiments in big labs because they don't get all the AI money.

Training is expensive, but they also have more than 10 000 of employees combined and they cost a lot of money.

I think at this point any attempt to prevent it will fail. AI code is almost everywhere, even in projects that were against it 6 months ago.

Historically, a lot vulnerabilities in PHP projects were because of PHP itself. Things like register_globals, remote includes/reads using functions that supposed to read local data and so on.

They would not catch it, it requires combining multiple vulns.

Yet the most viewed questions are about how to undo git commit, how to sort an array, how to select stuff in jquery, how to group by in SQL and so on.

General questions about programming languages, SQL and git don't change that much.

The graph shows questions, at some point they saturate, 90% of the stuff is answered. New frameworks contribute to the majority of the new questions.

Yeah, I hate the LLM wording too

> But there are also really good ones - so I can't afford to not take actual look at each, but it gets tiring and we need a solution.

There is not easy solution, the landscape has changed and if care about security, you just need to allocate more human resources. Another layer of LLM checks won't help.

Google Chrome used to report visited pages back to Google, not sure if this still the case. Also, Google Analytics can see visited pages and Google uses it.

Finding domains is easy, everybody uses CTL to find them.

The real price is how often a model uses subagents that scan your whole repository multiplied by thinking level.

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 40 malicious Python packages.

Unlike the LLM approach, my library is not susceptible to prompt injection and deobfuscates Python code. Where LLMs see "obfuscated code, potentially harmful", my library decodes it and sees what's happening inside.

https://github.com/rushter/hexora

It's not idiomatic implementation in Rust, just translation from C. A few Rust macros simplify implementation and cut the code in half.

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/14ffab7d31a31e5ab667...

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.

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.

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.

x4 pro will come with backlight and touch screen

https://www.youtube.com/watch?v=REKdjxYtMrI

It's the same on any package index now.

> What does any of that mean in practice?

They want you to spend more tokens

It’s a slop article based on a tweet.

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.

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.

Exactly, I also prefer when neighboring cells are automatically marked when you place a star

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.

I wish this article was not LLM written

It's used in ClickHouse as well. CH supports all known compression algos and they are documented pretty well.

> 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 attribution.

You don’t need to connect to duckdb, it’s just a process that you spawn.

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 databases at once, and so on.

And you get all the benefits of a database when you need to join or postprocess data from multiple sources.

https://rushter.com/blog/clickhouse-data-processing/