Also, no I didn't write the TODO, that has been there 13 years

For those that didn't follow...

This ProximityInfo object has different behaviour for gesture typing and tap typing.

To get it to swap behaviour, you pass either
1 for tap typing encoded as MAX_POINTER_COUNT
or
2 for gesture typing encoded as MAX_POINTER_COUNT_G

You might think that the "G" stands for gesture? Nope! It stands for "geometric"... but yes, it is for gesture typing despite that.

The abstractions, they be leaking captain. We gotta get to bailing.

So, the AOSP JNI lib has these fancy interfaces all over the place. There is one implementation for tap typing and how that interacts with their dictionary prefix tree, and there is an unimplemented one for gesture typing.

When user input is passed into the lib there is a ternary statement that picks either the gesture implementation, or the tap typing implementation.

Then, there is this other thing called "ProximityInfo" that doesn't know about input styles in theory.

in reality...

Have you heard about the artcraft proprietary software clones?

I think the idea it is a "clean room" is laughable, given I practically guarantee the source for the software they are cloning has been hoovered into at least one of the models they use.

One thing I hope catches on from this a programmatic interfaces for every action, and regression testing file outputs.

I want your prediction; I think this is where LLM boosters either prove it or flop. I don't really like it, but it may happen.

If someone genuinely, provably, does a perfect byte for byte decompilation of any arbitrary binary I can give them, but now it's magically written in rust no matter what the unknown source language was of the original binary, I'll pay for an LLM subscription. No joke. This would be the first thing that actually sounded useful to me.

I'll be waiting.

LLM boosters that don't know how compilers work are now claiming that you can do byte for byte decompilation by brute force guess and check now?

There are so many problems here I just can't even, nor can I odd.

The only way this works is if you have a compiler capable of reproducible builds, and it's the exact compiler originally used.

This is not what I'm hearing claimed... I'm hearing claims of rewriting arbitrary binaries made by unknown compilers by having an LLM write rust.

*sigh*

Serious question, is there any precedent of people who have tried to sue the various tech companies for collecting data on them, using typical recording consent laws? (1 party / 2 party whatever)

If someone can't sign a contract for you, certainly they can't sign the EULA for you with regards to their phone, their TV, or smart speaker, etc etc?

That person consented, without thinking, to the panopticon, but not everyone they facilitate passive recording of did.

#privacy #showerthoughts

Anyone happen to know much about c++ in #AndroidStudio?

Every time I change any c++ code (even a comment), 8 threads max out for the next 2 minutes. So, as I write doc strings I'm having to be next to a jet engine, when all my fans max out continuously.

Maybe there is a setting I can turn off? This has been driving me nuts. Honestly considering working in a dumb editor and only opening android studio once a day...

I assume it's trying to maintain the AST in the least efficient way possible.

What is even the point of rush shipping? I order economy and it shows up the next day. Can I ask for like economy-minus? I want to be like "no actually it's fine, I don't need this for like a week or three, take your time. Let the employees have bathroom breaks actually. Also, don't fly my package, it can go on a train."

@whitequark Hey, I just wanted to say thanks for working on #chipass! Also thanks to the others that aren't as active on my timeline.

I noticed it came to the repos in my distro very recently and I missed it! Immediately upon opening it... oh I love you all did with the qt6 work... My theme works!

Also, love that it looks for keepassxc settings? All my stuff works as I expect!

I know stuff like this is a thankless job most of the time, so let me personally thank you!

I do think there is a possibility of things improving here.

I think there is a very deep issue of how we as a society handle people that choose to benefit themselves at cost to others in ways we deem acceptable.

The problems around the normalisation of undermining one another's worth, casually, through means of proxied and obfuscated harms, I think is something people are slowly coming to grips with.

Not sure it will fully realise in my life time though.

I find stuff like this, honestly kind of distressing.

I don't fully understand how one can so easily think "my point in favour of humanity, and against plagiarism, is best made by sharing an arrangement of pixels that embodies disrespecting the work of other humans through mass plagiarism"

It's like watching a never ending stream of people willfully misinterpret obvious critiques, portrayed though media, as endorsements. (EG people who think 40k would be awesome)

Saw this in a comment about AI in mathematics. Was posted in _favour_ of human creativity.

A large number of people don't consider things they aren't practised in to be within the domain of meaningful creative tasks.

When one will slop one task, but not another, they must justify the idea that only the thing they are trained in matters. Rarely do two people agree on what can be cleanly delegation to the confabulation box.

Overcoming this is where solidarity lies.

So, this basically means, the prisoners dilemma is playing out constantly for people prompting genAI.

Every time someone shares something it has utility to them, but negative utility to everyone else... and if everyone else does to them what they just did, everyone is now worse off than when they started.

It is sort of like a weaponisation of digital post-scarcity.

I think most people don't actually like genAI output. However, their preferences for specific intersections of interest can override that dislike enough that they will like what was output, if _they_ prompted it.

They won't extend that same grace for genAI output, if they know it is such, when _someone else_ makes it.

This means, that every time someone prompts a thing, and puts it on the internet... the internet gets better for them, but to everyone else it got worse.

Also, the original quote has all interesting history.

https://en.wikipedia.org/wiki/The_death_of_one_man_is_a_tragedy,_the_death_of_millions_is_a_statistic

Spying on one person, in order to learn their habits to better separate them from their money is a crime.

Spying on a million people at the same time, to learn their habits to better separate them from their money is a business model.

Well, that is all the gains I am going to try for, in #drift_type during the grant.

I think that is pretty good, considering I was aiming for 60%

The problem is, we need a much faster dictionary to get the execution time back down. Right now, it is taking _way_ too long, by about 25-50x, is my guess.

It is way too slow to go in #heliboard.

That speed gap won't close until I have time to implement a _much_ faster dictionary. I also can't do that until I integrate some stuff from other algos.

This doesn't even take into account when something is being a pain to get through, and I am stressed for time.

I have seen a lot of people agree to things out of frustration. I have agreed to things out of frustration.

I would consider this agreeing under a form of duress, honestly.

In this category I would include those gdpr cookie banners, _especially_ when you have to deselect every "optional" group.

If your method of getting people to agree is harassment via UX flow, that is opt-out.

RE: https://mstdn.social/@theeclecticdyslexic/117089595149742886

This surprised me!

I am firmly in the "opt-out" camp, and I did not expect the "opt-in" camp to get above 10%.

On your worst day, could you accidentally forget to untick the box? Could you hit the brighter button? Could you miss the tiny x?

For me, it's yes. I have accidentally agreed to non-trivial things in interfaces. Now, imagine someone less tech savvy.

If, the purpose of a system is what it does...

This gets more people to agree when they don't mean to.

Therefore, it is opt-out.

🧵