rkwz

> AGPL is open source right? ducks

Out of the loop, is it considered not open source?

> Maud allows much tighter/automatic integration in this regard

Interesting, can you share some examples?

Yes, html is one of the unique languages of our era. It's very forgiving and lenient.

> This also explains why there are so few alternative browsers. Writing a new browser that supports everything that happens to work in the major browsers is ridiculously hard.

IMO one of the reasons Chrome won was because they supported both Firefox and IE's interpretation of web standards.

> As far as I know that's technically invalid HTML. You can't have an <a> as a child of a <button> (or even the other way around.) I'm also ignoring the invalid type="submit" against the <a> tag in the example button component in the post.

It definitely feels weird to have anything other than text nodes inside <button>, didn't realize it's actually invalid HTML.

I used this approach mainly to have the button component behave as a link button and a form submit button. Here's the actual code from my project - https://github.com/sheshbabu/mouji/blob/master/commons/compo...

What are the drawbacks of this approach? I'm pretty sure I've used this before, can't recall any downsides.

I apologize for the confusion. I meant React for component composition in codebase rather than for building SPAs.

https://news.ycombinator.com/item?id=40491188

Thanks! Go makes it really easy to go minimal dependencies.

I was able to build a website analytics project with just 2 dependencies (sqlite driver and bcrypt package) - https://github.com/sheshbabu/mouji?tab=readme-ov-file#philos...

> There eventually has to be JavaScript somewhere, and once you have to use that for any dynamic frontend, might as well use a proper JS/TS framework.

Yes, totally understand it's hard to escape JS in Web frontend. Even writing a simple dropdown requires a little bit of JS.

> Not being a frontend person first, I have learned to choose whatever they are using these days. It will be hard to hire, onboard, and maintain anything that strays too far for the normal patterns they use.

Since this is for my hobby project, it's more fun to be a irrational ;) I use standard React at work.

> Next has skyrocketed in popularity for a reason. It makes the React DX much better, losing a lot of the config hell from other framework/library amalgamations

Need to check Next again, I tried it many years back and it was good.

Yes, it's more about organizing your UI code/templates similar to a React codebase than running Go code in browser.

I think apart from vDOM, a big contribution from React is how we organize code.

I love React too, but it pulls in a lot of dependencies and the dependencies need periodic maintenance.

Sometimes you get the feeling that you're spending a significant chunk of your time not writing code but fighting with configs and dependency churn.

Using Go's html/template definitely makes the UI feel dated because of re-rendering, but I rarely notice it in my side project. My next exploration is to use HTMX.

I'm planning to gradually add dependencies like HTMX to see how far I can go to reach React's smoothness.

Ah yes, maybe my mistake!

> I noticed that the syntax highlighting for the Go html/template examples in the article is incomplete.

Sorry, it might be something wrong with my blog setup. I use hexo, might need to play around with the config.

I use this VSCode extension for coding and it works fairly well: https://marketplace.visualstudio.com/items?itemName=jinlimin...

Here's the link: https://www.sheshbabu.com/posts/react-like-composition-using...

Didn't realize HN doesn't convert link text to clickable links

Thanks! I wanted to use standard Go html/template to see just how bad the re-renders are, and then decide to use htmx. After building the project, I don’t mind the re-renders much as the latency is around 50ms and I don’t see any jank. I might introduce htmx in future

I had the same problem - really liked the way I used to do component composition in React, it was hard to go back.

Eventually found a way to do something similar using just the standard Go html/template.

I’ve written about the implementation here:

https://www.sheshbabu.com/posts/react-like-composition-using...

Congrats on shipping!

I'm curious how the search results rankings work, doesn't look like it's based on date or number of citations, but seems to be deterministic (persists over multiple searches). I did a keyword search using one word.

Wow, can you share more about the Playwright+VNC combo?

I use JMeter via Taurus, it displays visualization in the terminal.

My notes on how to use it - https://www.sheshbabu.com/posts/taurus-writing-jmeter-load-t...

> There's a big catch because often someone skilled enough to answer these questions would be better served actually doing things, e.g. writing code, and the people who can fill out these questions but not skilled enough to do the actual things are a weird middle-ground of mediocre that is hard to find. Really big companies often solve this by just paying overskilled people to do this for a few years, which is expensive and soul-destroying for the skilled person.

This is a really insightful comment, how do companies get around this issue (apart from paying overskilled people until they burnout)?

I interpret it as - A developer who can work with pre-packaged ML/LLM models that are offered by cloud companies

> relatively high barrier of entry compared to other software tech

Sorry, what do you mean by this?

Thanks for your work!

First time hearing about hyperscript, was curious why a new DSL instead of just using JS:

https://hyperscript.org/docs/#history

Sounds interesting, can you share a link to video if available?

> How do I script it? OF's automation is a must-have, as that's how I integrate it with my other apps. Extensive Shortcuts support (ala Things) is a minimum. AppleScript or JavaScript is better.

What's the usecase for scripting in todo apps? Are you using it like some kind of "cron job"?

> I use it for RAG and as a fancy knowledge extraction engine

Curious, can you share more details about your usecase?

Thanks for the notes, I thought it would be a more detailed prompt :) Any reason why you choose Opus instead of other LLMs, was it because of the 200k context window?

Interesting project! The instructions look similar to cucumber/gherkin tests but without the underlying instructions. Is the goal to automate navigation of arbitrary websites?

Thanks, this feels like Tauri but with Python (correct me of wrong).

Hoping to give it a try soon!

Since browser cache is no longer shared, why not self-host fonts for even better privacy?

Feels similar to Lindows - https://en.wikipedia.org/wiki/Linspire

> I was particularly interested in testing models’ ability to reason (i.e., perform a somewhat complex task that requires high context understanding) about out-of-distribution (i.e., unseen) data.

I was under the assumption that finetuneing LLMs was useful only when you need to change the model's tone (speak like a pirate, voldemort etc).

Are there other examples where LLMs were trained to reason a particular way?