3D printing experiment: I designed a little adapter to add a flashlight to my Swiffer. I call it the "Swiffer Explorer" now.

And it is fixed!

There is an issue tracked for it now:

issues.chromium.org/issues/45720...

And this seems to be a suggested fix:

chromium-review.googlesource.com/c/chromium/s...

Interesting to see how this takes its course.

The Chromium team is now aware of it:

issues.chromium.org/issues/40101...

Holding my breath for it to be solved as multiple of the web-based tools that I use daily depend on it.

Did a recent Android update break the File Access API in Chrome? On Android 16 with Chrome 142.0.7444.48 I get:

"Failed to execute 'showOpenFilePicker' on 'Window': The user aborted a request."

Google's own FSA demo also stopped working:

googlechromelabs.github.io/text-editor/

I see. Yes, I have built similar tooling for my workflow.

I'm not quite sure what it does. From skimming the readme, it looks like it expands a prompt with "Write a file in the language ... that does the following ..." and forwards it to an LLM?

True. But does the XMLHttpRequest solution have any downsides?

The File System Access API that is already supported by Chrome on desktop and mobile should be supported by Safari and Firefox as well.

It finally lets us write web apps that do not need to store the user's data in the cloud and lets the user backup their data themselves.

Fetch has all the tools you need. You can get the content length from the headers and the content transferred so var via read():

www.gnod.com/search/ai?q=....

Today I felt inspired to write down some thoughts about the current state of LLM-based coding and about a prompt coding experiment I am conducting:

www.gibney.org/prompt_coding

Theory that explains the p-values:

The Vitamin D capsules tasted more like Theanine to you than the actual Theanine capsules. And thinking you took Theanine improved your mood.

Chart that shows China at over 9000TWh while every other country is below 5000TWh.

In the era of AI, the intelligence of a country will be equal to

electricity * chips * software

Not sure how it will look on the chips and software front. But in terms of electricity, China seems to be hard to catch up with.

I'm interested in the question if a Surface Pro (With Snapdragon X Plus, which is an ARM chip) makes a good dev machine. As far as I know, you cannot boot a stock Linux distro on this device.

As for VM, which one would you suggest to run on an ARM machine?

Hmm.. reading through that thread, it sounds like they did not use the Docker version that came with their distribution. I wonder if the version that came with the distribution worked better.

The ticket is about the Windows version of Docker. Does it really apply to the Linux version?

When you say "some Docker stuff", do you mean you typed "docker run someimage" in the terminal inside WSL2 and it did not work?

I see. What's a weird thing you experienced?

Would you say WSL can replace a native Linux machine for web development? So Apache, Python, Vim etc run in WSL?

The html-editor with instant preview is still my favorite tool to do HTML/CSS/JS experiments on the go.

no-gravity.github.io/html_editor/

So I added a some more features today to make it more convenient.

With the json arrow functions, SQLite makes a nice document store. If you have a json "book" column, then selecting all book names is just:

SELECT book->>'name' FROM books;

But I miss auto increments.

Without an additional 'id' column, inserting a new book with a unique id is tricky.

I started adding additional infos on the www.productchart.com item pages when hovering the data fields with the mouse.

The iPad Pro has a nice aspect ratio of 1.33. Most tablets are at 1.6. I think the more squarish, the better for coding.

It is also nicely lightweight at 578g for the 13" version.

Isn't Waymo already doing over 150k autonomous rides per week?

Oh, I see!

What is an "order 6 Mandelbrot set"?

Experimenting with a tablet based development workflow.

#buildinpublic

From a usability standpoint, the AT Protocol is really nice. You prove ownership of a domain simply by placing a text file containing your DID string on it. The domain then becomes your handle.

I tend towards the first one.

If Google is not great at sorting results, that would have a few interesting consequences.

The first thing that comes to mind is that it would be interesting to compile a list of tools that are the best tool for "xyz" by a good margin, but are not listed on the first search result page for "xyz".

In the end, Google always thinks what users think. I have yet to see a tool compelling enough for users to bookmark and regularly use that Google does not rank.