Reading this article, I find it pretty thought-provoking, everyone: AI is removing the middle class of software engineering.
The author argues that artificial intelligence has now removed the speed limit on software development. I don't know about you, but I really do feel that's the case, because these days, if you ask me to sit down and type out code line by line, it's honestly... very difficult. Instead, I find myself doing more of the work of a problem solver, reviewing and evaluating things. This is both beneficial and harmful. The benefit is speed, while the downside is feeling like I'm able to accomplish a lot, when in reality that may not necessarily be the case.
Large language models, AI Agents/Harness... are now changing the hierarchy of people working in the field. They help new programmers work dramatically faster, while those at the senior level are able to fully leverage their power based on the years of experience they've accumulated. So what do you think will happen to the people in between those 2 levels?
Finally managed to carve out some extra time to revamp the website's homepage 🥳
Did you spot what I was going for 🤓? What do you think of it?
Hơi rối a ơi, chắc a muốn đẩy tương tác à?
Đúng rồi á bạn, mình đẩy mục Threads lên đầu vì nó được viết thường xuyên 😅. Còn bạn thấy rối chỗ nào thì chỉ mình biết thêm với nhé!
Here's a quick article explaining what Harness is: Harness – definition by the Cambridge Dictionary
Many of you have probably come across Harness or the term "Harness Agents". How is it different from AI agents, AI, or agents in general? Just give the article a read and you'll get it right away.
On the same day, Qwen and Z.ai introduced 2 Flash versions of their current top-tier models: Qwen3.8-Flash-Next & GLM-5.3-Flash.
Not sure about the quality yet but the pricing is very attractive, even beating DeepSeek Flash 🥶
Vậy ra Ox alpha - mẫu model miễn phí trên OpenRouter 1 tuần qua chính là GLM 5.3 Flash
So basically, right after DeepSeek raised their prices, a whole 'token crisis' broke out, you guys. For the first time, I completely burned through my OpenCode Go plan. Now I'm stuck scrambling around looking for alternatives.
Part of the problem is that OpenCode has been acting up lately. There are bugs everywhere — from the CLI to the API. It's super frustrating to use. Any session that runs a bit longer is guaranteed to hit some error, not to mention it burns through tokens like crazy. There was even a time when I was doing just one simple task and it ate up my entire 5-hour limit — and I was on DeepSeek Pro V4 at the time.
Now I'm thinking of a little hack: using CC Switch to spin up a Proxy server and plug it into Claude Code as a temporary fix. The first goal is to replace OpenCode's CLI, and the second is to take this chance to test out whether other models actually work well with Claude Code.
Every now and then, another Coding Agent tool pops up. Vercel recently released fx.sh — a lightweight CLI tool written in Zig, barely over 6MB.
You'd expect people to praise it, but instead many are wondering why something this simple needs 6MB 😌?
That's actually how I stumbled upon hax — a coding agent written in C, just a few hundred KB 🫣
Earlier, many people were really into hunting for leaked System Prompts from some major systems. Evidence shows there are GitHub repositories with tens of thousands of stars. Of course, Anthropic with Claude wasn't an exception.
But recently, Anthropic publicly released the System Prompts they use for each model family. They're not as long as what we've seen online, but they contain so much interesting info that it's... amusing 🤓
Do you guys remember how Anthropic watermarks content generated by Claude? According to their explainer on how it works, they firmly claim that AI-generated content won’t be altered. However, whether that’s actually true is something the article Anthropic’s ‘Watermark’ Text Adulteration in Claude Is a Perversion of Writing helps clarify for us.
At first, when I read it, I did feel a bit uneasy because the algorithm they described is based on the frequency or combination of words inserted into the text to tell it apart, and inserting extra words or using different wording is already enough to change the content to some extent, whether a lot or a little. After reading the article, I could see just how strongly the author reacted to Anthropic’s way of putting it.
Sorry for the inconsistent posts lately, I've been learning Rust and it's so fascinating that I haven't had the focus to write. I've had a series of blog posts about my journey learning Rust (from about 2 years ago), but they all stayed at the theory level. It wasn't until I actually started coding that I realized how challenging it really is
Qwen3.8-27B has just been open-sourced by Alibaba. It is currently attracting a lot of attention online. Many people say that version 3.8 "thinks" more deeply about problems, so its response time is longer, but it delivers noticeably better results.