mg

Well, sure - if someone wanted to be the hoster for all mail traffic on planet earth, then they would need a lot of resources and a business model to finance it.

But if that is their plan, why would they have used ActivityPub in the first place?

So I guess they could just shut their door as soon as the traffic is too much for them to handle. And tell people to look for another host or host themselfes.

Isn't Pixelfed using the ActivityPub protocol?

As far as I know, that makes it scale like email. When you post something, your host sends it to all recipients - everyone who subscribed to your posts.

That does not sound like a lot of traffic. Even if one has 10k followers, as the post would only go once to each of their hosts. So if the 10k followers are on 100 hosts, that means 100 messages.

I think this is where the future of coding is. It is still useful to be a coder, the more experienced the better. But you will not write or edit a lot of lines anymore. You will organize the codebase in a way AI can handle it, make architectural decisions and organize the workflow around AI doing the actual coding.

The way I currently do this is that I wrote a small python file that I can start with

    llmcode.py /path/to/repo

Which then offers a simple web interface at localhost:8080 where I can select the files to serialize and describe a task.

It then creates a prompt like this:

    Look at the code files below and do the following:

    {task_description}

    Output all files that you need to change in full again,
    including your changes. In the same format as I provide
    the files below, that means each file starts with
    filename: and ends with :filename
    Under no circumstances output any other text, no additional
    infos, no code formatting chars. Only the code in the
    given format.

    Here are the files:

    somefile.py:
    ...code of somefile.py...
    :somefile.py

    someotherfile.py:
    ...code of someotherfile.py...
    :someotherfile.py

    assets/css/somestyles.css:
    ...code of somestyles.css...
    :assets/css/somestyles.css

    etc

Then llmcode.py sends it to an LLM, parses the output and writes the files back to disk.

I then look at the changes via "git diff".

It's quite fascinating. I often only make minor changes before accepting the "pull request" the llm made. Sometimes I have to make no changes at all.

If you have an application that does not support RSS/atom yet, you can easily add support with a few lines of Python.

Or use an open source module. Here is a general atom feed generator that I wrote and published under the GPL:

https://github.com/no-gravity/atomfeed.py

Just 14 lines of Python. And it has been reliably serving the feed for my own website for quite a while now.

Would this also be the best option if you just want to convert plain text files to audio?

I find it an interesting question to what degree the design decisions of WordPress are still the right ones in 2025. A PHP backend and frontend running on the same machine, serving dynamically rendered content from a MySql database.

Personally, I prefer to have my content in text files, which are using a DSL similar to Yaml but optimized for use in a CMS. And render it to static html files via Python. A process that happens on a different machine. The html files are then pushed to the server that serves the frontend.

Because most applications have 27 active users per day and a $10/month VPS can handle 100000.

You can already do POST requests by embedding forms and/or JS. And with the proposed <button method=DELETE> you could also embed that. So I don't see how the proposal of adding more HTTP methods to html elements prevents abuse.

GET because that is also the default for all other elements I think. form, a, img, iframe, video...

Ok, but what is the advantage to be "clear at the http level"?

What is the upside of

    <button action="/users/354" method="DELETE"></button>

over

    <button action="/users/delete?id=354"></button>

?

Hey, this is pretty cool!

Have you considered publishing it under an open source license?

Then I could see myself working on some features like: Selectable color palette, drag&drop and pinch-to-zoom on mobile and fractional rendering (so that when you move the position, only the new pixels get calculated).

I have yet to see a Mandelbrot explorer written in Javascript that allows infinite zoom without losing detail and a good UI that works on desktop on mobile.

Does anybody know one?

If there is none, I would build one this year. If anyone wants to join forces, let me know.

If it is triggered by the customizations you did in Firefox, then running a fresh Firefox in a container might help:

    docker run -it --rm -e DISPLAY --net=host -v $XAUTHORITY:/root/.Xauthority -v /tmp/.X11-unix:/tmp/.X11-unix debian:12-slim

Then inside the container, run:

    apt update
    apt install firefox-esr
    firefox

My biggest surprise in AI 2024 was that it seems that Bing did not manage to take any market share from Google:

https://bsky.app/profile/gnod.bsky.social/post/3ld43usdckk25

ChatGPT now has hundreds of millions of users. Personally, I do over 50% of my "searches" on LLMs now. And Bing, while having access to ChatGPT and being the default search engine on every Windows machine seems to not have gained any traction from it.

What is the core of the problem?

Is it that when you have two colors like RGB(0,0,255) and RGB(255,255,0) there is no easy formula to calculate what a mix of these two colors would look like in the real world?

And therefore one needs a lookup table? And since 16Mx16M is a large number, a complete lookup table is impractical?

Not sure the author's main point - that computers become slower and less reliable over time - holds.

HN is snappy and rarely fails to function. I can't remember ever using a better group communication tool. The author's website is also snappy and worked perfectly to let them express their view and for me to read it.

A lot of software I use today is software I wrote myself and that works in the browser. It is reliable and snappy too. And I write more and more of it with the help of LLMs. Just yesterday, I wrote an image editing tool with a very special feature set and was amazed how much faster I was able to implement features in perfectly elegant code via this approach:

    1: Give the existing code to an LLM and instruct it to add a feature.
    2: Test the feature to make sure it works.
    3: Go through the changes via git diff in one window and clean them up in another window in VIM.
    4: Test again.
    5: Commit.
    6: Goto 1

This way, the development of this new tool was multiple times faster than if I had written it manually.

To me, the main questionmark about the future of computers is how browsers will develop. The one thing I am missing is that currently no mobile browser supports the File System Access API. On desktop, Chrome does. And that allows users to securely safe their data on their own device. So web software does not have to be like "All your data needs to be in the cloud". Chrome already works on bringing that to mobile. Then the current state of affairs on Desktop and Android will be perfect from my perspective. And all I hope for is that Safari follows. So that iOS users can comfortably use my software too.

Wow, sandbox="allow-scripts" looks pretty good. Doing some additional research, it seems one can disallow all network requests via Content-Security-Policy directives:

    <iframe sandbox="allow-scripts"></iframe>

    <script>
        document.querySelector('iframe').srcdoc=`
            <meta http-equiv="Content-Security-Policy" content="default-src 'none'; script-src 'unsafe-inline'">
            <span>Hello</span>
            <script>
                document.querySelector('span').innerHTML+=" world"; 
                fetch('/');
            </sc`+`ript>
            <img src='https://placecats.com/300/200'>
        `;
    </script>

The JS runs, but no network requests are made. I wonder if that solves all headaches regarding the execution of user provided JS code?

What is the state of JavaScript sandboxing these days?

I would like to build a site geared towards this kind of algorithmic art. Where one can paste the JS code snippets and see the resulting image. Similar to my html editor with instant preview:

https://github.com/no-gravity/html_editor

But with the ability to link to the code.

Is it feasible these days to build something like this, but without the ability of the code to jump out of the result frame? Like no external http requests, no fiddling with the window etc?

Is that really cheaper than to just use a VPS for $5/month which offers compute and storage and everything?

The article says this about VPSs:

    You often end up paying for
    resources even when you're
    not using them.

But how is that different with S3? They also do not store the data for free.

And you can delete a VPS via API just like you can delete an S3 container.

Yes, tablet.

A new iPad with 13" would probably be the lightest external monitor with battery one can get.

Sure, those are full computers but you can't run Linux on them.

No mention of the weight?

If it is lightweight enough and could be powered by a table which at the same time functions as the monitor, this would be a dream. Finally a mobile Linux solution.

Create videos in the browser which the user then can download and play on their device.

The problem withe the MediaRecorder API is that it saves the current timestamp with the frames. So the video plays at the speed it is created. Therefore you can't use the MediaRecorder API for video processing. That's why I referenced the VideoEncoder API in the prompt.

Yay, I just tried it on my iPad and it works!

When you say "GPT-o1", do you mean the model "o1-preview"? Because I think that is the only o1 I can access via the API.

Is there an LLM that can solve the following simple coding task?

    Make a simple HTML page which
    uses the VideoEncoder API to
    create a video that the user
    can download.

Since the VideoEncoder API is made for this exact use case and is publicly available, it should be able for an LLM to figure it out. But I have yet to see an LLM answer with a working solution.

Currently, this is my favorite test prompt for AI coding tools:

    Make a simple HTML page which
    uses the VideoEncoder API to
    create a video that the user
    can download.

So far, not a single AI has managed to create a working solution.

I don't know why. The AIs seem to have an understanding of the VideoEncoder API, so it seems it's not a problem of not having the infos they need. But none comes up with something that works.

We have to add the deprecation and maintenance of the car.

Plus I guess they need high resolution maps? Not sure if that is a significant cost factor.

    our service – which already provides over 150,000
    trips per week across Phoenix, Los Angeles, San
    Francisco, and Austin

Interesting. That's about 8 million rides per year.

I wonder how close they are to being profitable? As soon as they are getting close to being profitable, they will probably scale this up super fast.

I don't know how much Google invested into Waymo so far. Something like $10B?

If they at some point make $10 per ride, they would only need something like 50 million rides per year to justify that investment with a p/e ratio of 20.

To go from 8M rides to 50M in 5 years they would have to increase their capacity by 50% per year. Might be possible?

Related: Why pipes can be indeterministic.

https://www.gibney.org/the_output_of_linux_pipes_can_be_inde...

That is what I mean with "does one thing". The author made the right choice to put just one feature into this repo.

If you publish a repo with many features, bells and whistles, it is lost on me.

If you publish a repo that aims to do one thing in the best way possible, I will take a look at it, might use it, might contribute if I stumble across something that can be improved.

Repos like this should stay as small and focussed as they are. New features should go into new repos that make use of repos like this one.