eclexic

https://mstdn.social/@theeclecticdyslexic

Please boost this toot!
https://mstdn.social/@theeclecticdyslexic/116169411669556587. Even though it's old, it's important! My job right now is to replace gesture typing in #Heliboard; It helps to have more people contributing data, or spreading the word around!

I'm an Industrialised information consumption apparatus, FOSS fanatic, misguided believer in a better world, and owner of too many opinions.

If you follow me for Heliboard stuff, send me a message I will follow you back.

pronouns: he / they / any (...except "it")

Public Key

npub1t9neskuk8y5z9k7zxpj8yr9dqds6drzy7jjwht2jyk2qqqpje3qq2pnd33

NIP-05 Address

theeclecticdyslexic_at_mstdn.social@momostr.pink

Profile Code

nprofile1qqs9jeuctwtrj2pzm0prqerjpjksxcdx33z0ff8t44fzt9qqqqevcsqpz3mhxue69uhhyetvv9ujuerpd46hxtnfduqs6amnwvaz7tmwdaejumr0dslmna66

Publishing to

relay.damus.io

nos.lol

Author Public Key

npub1t9neskuk8y5z9k7zxpj8yr9dqds6drzy7jjwht2jyk2qqqpje3qq2pnd33

Show more details

Last Notes

2026-10-08 07:03:11 UTC

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*

2026-10-05 22:08:35 UTC

- reply

how do they manage details? I've been looking for some speakers I could put on my desktop, but I often listen to audio at increased speed, and certain speakers can't handle it. If they are too muddy, the sped up words blur together.

2026-10-05 03:32:33 UTC

- reply

I do agree with you. I don't think this is women's job at all.
I just mean to say, that men keep being given the vision of how to have meaning... and quite regularly that vision takes the form of being needed by a woman.
In an egalitarian society, where a man isn't needed in order for a woman to live a stable life... this model breaks down. These men need to find another way to find meaning, but good options are not being readily presented to them.

2026-10-05 03:19:02 UTC

- reply

My opinion on this might be a little weird?
I think it is clear there is a crisis of meaning among young men, and sexism is also the air we all breathe. I think those are bidirectionally reinforcing, but it _mostly_ goes from latter to former imo. One simple way to feel needed is to disempower people you want dependent on you.
Until the need to dominate as a path to meaning is addressed, this problem will continue... and we will keep flirting with regressing to women as property.

2026-09-30 08:04:45 UTC

- reply

Sorry, I am kind of basic when it comes to tabletop RPG shows. I have only watched critical role and stuff from dropout.
What parts of tabletop RPG shows do you like the most? Maybe knowing that could help tailor suggestions? I am sure there are some pretty crunchy shows out there, and some very RP driven shows... and among the RP shows, there would be plenty of variation too I am sure! Yes and yuck 'em ups, story driven, and character driven stuff!

2026-09-27 20:35:53 UTC

- reply

yes, this is really common in basically everything.
Only the people trying to do better get judged, because they are within reach of people with no power.
However, the people who really are the problem get a pass, because people don't think they can do anything about that.
You'll get judged even if you do everything perfect, because you won't have done enough.

2026-09-27 20:17:33 UTC

- reply

Schrödinger's slop

2026-09-24 23:49:38 UTC

- reply

classic. Nothing Canadians hate more than being asked to vote before a term is up.

2026-09-24 07:15:57 UTC

- reply

I agree mostly.
I think this particular elevator looks like it could be going down to a floor most people haven't seen before though. That can be scary, especially if you are maxed out but still unable to meaningfully prepare.
One part I don't agree with is there are some metrics by which things are actually getting worse with successive crises faster than they can improve between them. We are missing some vital tools for the recovery phases.

2026-09-23 08:23:47 UTC

- reply

Ya, I mean. I don't think this killed the project on its own. It was a confluence of stuff for sure, but even debugging this took a good amount of effort. I think I could pick up the same project again when I have time. I still like the idea behind it.
It kind of amazes me that I am still getting emails from that thread. People are still walking head long into it, and then finding out after a big debugging session that this was an intentional choice that was communicated nowhere.

2026-09-23 07:49:58 UTC

- reply

Ya, for sure. I got around it. It just took a lot of the wind out of my sails. I didn't feel up to discovering a bunch more of these kind of surprises after that at the time.

2026-09-23 07:38:32 UTC

- reply

basically, you can't easily do controller rebinding to triggers without using what godot considers an anti-pattern in input handling.
Huge PITA. Took me so much work to get around this that wasn't at all necessary.

2026-09-23 07:35:36 UTC

- reply

https://github.com/godotengine/godot/issues/72636

2026-09-23 07:33:02 UTC

- reply

that sounds fun! I am not very familiar with these types of games, but I am really glad you have a vision that is solidifying!
I like godot. It isn't perfect by any means. There was one really horrible design decision they made with 4.0 that really killed some of my enthusiasm for the project I was trying to work on... They have pushed back fixing it for several years at this point. However, outside of that, I liked the experience of working in godot.

2026-09-23 04:43:12 UTC

- reply

cute art! I think I have missed your updates about against the goblin overking. Did you do the art?
Also, are you using an ootb engine or something bespoke?

2026-09-21 02:46:05 UTC

- reply

I introduced my partner to it, had a lot of fun going through the whole thing in order. I actually saw it out of order the first time. There are actually a lot of really good details I didn't remember, which is rare for me to find in a show!

2026-09-21 02:35:09 UTC

- reply

did you also recently rewatch?

2026-09-20 18:09:50 UTC

- reply

I can say confidently, layouts matter, but I don't pick them for speed. I used Dvorak for a decade to reduce the onset of an RSI, before getting something split, adjustable, and programmable. Much better options for that.
Reliably, the fastest typists I've heard of in competitions use qwerty, but if I had to use their form I would have to give up using a keyboard.
As far as speed goes though, no layout will ever beat steno anyway. Pick layouts for health reasons, not speed.

2026-09-19 03:19:16 UTC

- reply

can you pick the message, or is that just implied?
This also sounds like hell, because I remember I tried really hard at duolingo for a year, and got nowhere with it. I am not totally convinced it really works that well. It seems like a really ineffective way to learn. Maybe an okay way to brush up if you are rusty though?

2026-09-17 07:55:42 UTC

- reply

Ya but explorer had the distinction of being big enough to drag the whole internet through the mud for ages.

2026-09-17 07:06:48 UTC

- reply

early, but I wouldn't say the first. Vista was pretty enshittified and so was internet explorer.

2026-09-16 18:00:32 UTC

- reply

There was a point where the stop signs in Quebec said "arrêt"; in France AFAIK they have always said "stop".
I think this says something profound about the size of the stick up the Quebec government's ass.

2026-09-15 21:34:04 UTC

- reply

I couldn't even pick an answer here, because the framing is too far from the way I think about this.
Slowing down means little here. The development isn't my worry, it's the frenzied construction and disregard for the quality of life of people left in the wake of this "progress".
There is so much money being blown on something that has an negative impact on human quality of life.
I want everyone to imagine taking all this effort, and using it on transit, batteries, and renewables.

2026-09-10 20:14:05 UTC

- reply

there was a point, not long ago, where comic book super heroes were clearly our modern mythology. I think we, as a society, are starting to look for a new mythology.
I don't think comics are quite dead yet, but I think we are due for a shift. Comic book heroes are sort of the mythology tied to American hegemony. All the big heroes I can think of are quintessentially American propaganda.
I think as America's cultural dominance wanes, a new mythology will emerge.
Wonder what is next

2026-09-05 20:38:29 UTC

- reply

I don't think it was republicans that figured this out. I think it was an emergent property of the massive boat anchor on the modern economy that is attention monetisation.

2026-09-01 17:37:44 UTC

- reply

that's all true. But I don't have to give Carney credit when he is actively trying to strip the copper out of our democratic process.
I heard one person, that I think usually has pretty good takes, say they think that if the conservatives had a minority right now the damage would actually be less, because people would be pushing back on them, instead of acquiescing/cheering. I'm not sure if that's true, but I understand the logic.

2026-09-01 17:26:27 UTC

- reply

I'm less enthused with my vote, though I'm still okay with it I suppose. I think he is taking advantage of this to roll out some shock doctrine, and because liberals are the classical party of safety for Canada, he isn't getting called out on it, in fact he is being lauded for it mostly.

2026-08-29 00:31:57 UTC

- reply

what kind of phone? I saw someone mention in an earlier comment section they thought you had a folding Samsung? If it has a weird form factor give heliboard a try, we always could use someone testing things in an unusual form factors. If it isn't a weird form factor, give heliboard a try anyway, we are only on fdroid.

2026-08-16 18:44:10 UTC

I was thinking about the talk @npub1764…nuj0 gave at #fedicon. One point brought up was fedi users don't stay fedi users very long. They get an account, but the account goes inactive not too long after signing up.
iirc a large factor in this is they don't have anyone to follow. I was talking to someone with a different reason; they didn't get notifications on their iPhone, only to their email.
I do wonder how many people are lost to UI problems we don't even know about. Really difficult problem.

2026-08-14 22:00:11 UTC

- reply

A friend wanted to get an old game working for their dad that never works out of the box on windows. https://en.wikipedia.org/wiki/Myst
I told them to try on their steamdeck; they hadn't yet, expecting it to not work.
Download start to playing was ~1 minute.
They lent their steam deck to their dad for Christmas and then bought one for him later.
They gave up windows not long after, because it was clear Linux trends towards more stable with time, not less. (not universal but mostly true)

2026-08-13 18:23:53 UTC

If the path requiring the fewest clicks results in you agreeing to something, is that considered opt-in or opt-out...
(I have an opinion here, but I will share when the poll is done.)

2026-08-12 13:56:11 UTC

- reply

revving sounds in many pickup trucks are done with speakers, iirc. The actual engine doesn't waste enough energy as noise for people to feel powerful when accelerating.

2026-07-27 16:32:45 UTC

- reply

Did you print bases for the marker cards? We have some of those. They are printed so that the card stock the tokens are on holds tight enough into the base that you can pick them up by the card. It is pretty cool actually having our pathfinder pawns style tokens on top of a TV that has our battle map on it. Let's us have nice art on the tokens, and I get to use the software from https://www.megasploot.com/ to make our battle and world maps.

2026-07-27 16:00:39 UTC

- reply

these are some fancy toys. My group just uses a table I made out of a flat screen TV, and we only get maps when I have the time to draw a new one. 🫠

2026-07-26 14:45:40 UTC

- reply

I gotta agree. It's not capability. It's interest. Most people just don't really care to know. And honestly, they shouldn't have to know. The systems should be good enough they don't need to know. We are a long way from that, but that's how it _should_ be.

2026-07-15 08:17:28 UTC

- reply

sorry to hear that. I don't know what's in your feed, but I can say the stuff I've been putting on here sure has me excited. 😅
Hope you had fun with the cat watching!

2026-07-14 05:48:31 UTC

- reply

It's my opinion that people don't change settings and will use LLMs for mundane tasks easily covered by existing tools, because most interfaces are inscrutable to most users. They don't even know what they don't know, because discoverability is basically not a thing in most interfaces.
They aren't stupid, the interfaces are wrong. I don't know how to fix them, but they are wrong. It should not take the number of hours I have spent to learn to use a computer effortlessly.

2026-07-10 19:18:32 UTC

- reply

#Heliboard 4.0 just released on github; it will be on fdroid soon. 4.0 includes opt-in background data gathering!
Collecting gestures from daily use has some privacy risk; I made a video about how to leverage our privacy safeguards.
Boosts please!
PT: https://makertube.net/w/pPywMiF7kjumFfMrQKHzoU
YT: https://m.youtube.com/watch?v=VJLzVEhY2PY
Text (WIP): https://github.com/HeliBorg/HeliBoard/wiki/Background-Gesture-Data-Gathering
Active gathering was also changed to focus on common words. We still want active gathering data; background data is a supplement not a replacement!
2/3

2026-07-07 22:30:19 UTC

- reply

The point isn't perfect enforcement I would think. I would suppose, the point is to have a policy to point to when someone sends you a vibed PR, so that we don't need to relitigate it every time. You simply tap the sign. Plenty of rules exist for purposes like this.
While it may be "unenforceable" any reduction in the deluge is probably worth it, because right now it is a DDoS on many maintainers brains.

2026-07-06 07:16:11 UTC

- reply

in short, I think a personalised context and novel ways of connecting concepts through language is inexorable with saying things people will dislike.
To avoid this entirely is to speak, but say nothing.

2026-07-06 07:09:58 UTC

- reply

I think at a certain point... being disliked for no reason is part of existing in the world.
Avoiding this I think is part being precise in language, but it's also part removing the context of your own view point. This can also be done without an LLM, many politicians are extreme examples of this.
I think if you sand off enough edges, it replaces the linguistic associations you have in your mind for the ones averaged over society, and no longer encodes _your_ model of the world.

2026-06-17 21:18:55 UTC

- reply

the only genuine mistake I've seen that actually suggests genAI with touch ups is the corner creating impossible geometry with the lines of code sign.

2026-06-17 20:47:51 UTC

- reply

seems unlikely to me it's AI gen. The story points follow the Fibonacci sequence, all the text is the same font. None of the lines I see are converging in strange ways. Card 5 looks off, but that may be an attempt at portraying two different refracting lenses; they just forgot to do the same for the 21.
edit: the lines on the stands/glass back corner of the story points is far too correct. I really doubt it.
edit2: read the thread, I am _very_ wrong.
@npub1xq6…d8p3

2026-06-11 02:18:51 UTC

- reply

That steep? oof.
I mean it is the same here, last I checked it was _cheaper_ to fly than to take a train... and it would take less than a third of the time.
Better believe the flight is more resource intense, but it is subsidised like no tomorrow.

2026-06-07 19:25:41 UTC

- reply

oh jeez stux, I thought for sure you were covered, when I checked the donation pages.
How short are we currently? I can at least send a little.

2026-05-26 07:51:08 UTC

- reply

Anyway, thanks for two wonderful conversations Sasha. Have a good one!

2026-05-26 07:41:07 UTC

- reply

mmm, I just hope we don't end up degrading the quality of all these services, through substandard automation, so much that infrastructure which keeps people alive starts failing. That is my big fear. I think it is very possible.

2026-05-26 07:20:30 UTC

- reply

Only thing I think I will add is, I don't think that a lot of software dev jobs are getting properly replaced... but I also don't think it matters.
Right now, LLMs are acting sort of like the introduction of the gun. You probably would have been better off with a legion of longbowmen, but the line of guns is so much easier to assemble.
This poses a problem that is common in automation. The product may be worse, but the power afforded over labourers is worth it to capital...

2026-05-26 06:59:28 UTC

- reply

I think the "does nothing good" point is a real pain because it is often conflation caused by the marketing... where genAI wants to take credit for every tailor made ML victory.
"AI is stealing jobs" I think comes from software devs getting the old conservative jolt of "oh... when it happens to me now it's bad!"
Most of these I can only agree with in terms of scale. It is all the worst things we were already steeped in, but faster. I think we are accruing a lot of "culture debt".

2026-05-26 06:21:36 UTC

- reply

What are the worst ones you are hearing? I have plenty of complaints.
Most of them boil down to:
- chokepoint capitalism
- offloading work to someone else isn't automation
- destroying the commons as a business model
- slowing down validation is rarely worth speeding up input
- I don't trust anyone to do due diligence with it
- don't like the concepts of truth and chains of attribution being on fire
I probably missed one.
I am unsure how I feel about copyright being on fire.

2026-05-26 05:26:38 UTC

- reply

Fair enough. I meant more that they go completely unchallenged in computer science, not that they don't exist anywhere else. It was like a total monolith. I had never even met most of the women in computer science when I was in it at university. I honestly didn't know they existed, because they steered so clear of the dickheads. They had their own parallel student society.

2026-05-25 20:34:54 UTC

- reply

oh ya! I've been saying for ages computer science is the last bastion of the sexist dickhead. No argument from me. I will admit to having been one of them.
I'm thinking about my dental floss changing their process and now it snaps when I use it ~20% of the time when it was near indestructible 2 years ago.
Doctorow actually explicitly said, in the book, that he expected and is at peace with culture using his word until it's just a term divorced from it's original meaning.
@npub16y8…d455 @npub1ljm…w8tm

2026-05-25 19:53:09 UTC

- reply

Ya, I don't think I even understand. Is the desire to be rid of enshitification nostalgia, or is enshitifying happening because of nostalgia? I get the former, but read the meme as the latter.
Also, call me out if necessary but, I don't think it's wrong to be nostalgic for the immediate past when products and services I used were objectively better? Like within the last 2 years I can list so many things I interact with daily that have had their quality degraded immensely.

2026-05-16 20:07:53 UTC

- reply

I mean, even with ethical training I'm not convinced the way LLMs impact the priors of culture is net positive.
It seems to me it is currently most effectively used to DoS attack, or otherwise mislead, human brains. I think it has accelerated nearly every problem we currently face much more than it has accelerated the solutions.

2026-05-14 19:39:29 UTC

- reply

I was listening to something @npub1fdr…lvhs said recently, about the NLRB. He made the point that actually the NLRB was designed to put a violent labour revolution back in the bottle the only feasible way.
I think we might need to start thinking about copyright this way. If they are going to launder copyleft, maybe it is time to start laundering proprietary code that leaks. They have declared licenses dead, maybe we should believe them.

2026-04-16 23:11:28 UTC

- reply

Are you suggesting we shouldn't violate all copyright in the world as a treat? I don't understand. /s

2026-04-11 18:23:26 UTC

- reply

then save yourself on SaaS fees if you are using LLMs for this I guess, because people have already reproduced the bulk of what mythos did with existing open models.
https://aisle.com/blog/ai-cybersecurity-after-mythos-the-jagged-frontier
You just have to work around the context window size, basically.
Really, they are just trying to get a favourable IPO. You don't need to hold water for them, no matter who else might be.

2026-04-11 18:11:37 UTC

- reply

So, many of the vulnerabilities are in practise, not possible to exploit.
Probably good to fix, but zero days they are not.
https://www.tomshardware.com/tech-industry/artificial-intelligence/anthropics-claude-mythos-isnt-a-sentient-super-hacker-its-a-sales-pitch-claims-of-thousands-of-severe-zero-days-rely-on-just-198-manual-reviews
I would suspect the same happened with Greg. I think you should try to be more critical of claims coming out of these companies.

2026-04-08 06:05:15 UTC

- reply

thank you, I will read more about this. I still suspect this isn't entirely on the level, with this specific article. Also, hard to say if it would keep being good, even if this is accurate, once this round of vulnerabilities is found. It's training is based on bug reports that likely deal with near identical code in other places.

2026-04-08 05:53:04 UTC

- reply

which kernel devs? I only saw Jim Zemlin and other executives quoted.
I suspect they aren't releasing it due to inference costs. This is spin.
If this was true at face value, Claude code's own source would look _very_ different. I suspect the false positives are underreported. I bet they send a filtered set to cybergym. ARR is proof enough they cook metrics.
With the Firefox vulnerabilities, no false positive rate was given iirc. All 3 disclosures needed the sandbox disabled.

2026-03-21 08:05:27 UTC

- reply

I saw one that was very curious. It was a Markov chain bot. (Not an LLM, definitely old school Markov chains). It was also followed by another bot.
While the report was processing, it started spamming news stories, about 10 times an hour. Looked like it was probably paid to flood timelines with propaganda?
I think you guys need more auto flagging tools, and maybe a stricter admission process. Looked to me like state funded media disinfo. They are going to have more human hours than you.

2026-03-06 21:09:10 UTC

- reply

KDE user spotted? Stux, didn't know you were on the plasma side.

2026-03-04 06:09:08 UTC

#Heliboard is gathering gesture typing data, so that we can implement a #FOSS alternative to the closed gesture typing library Heliboard currently relies on. The data we collect will be released under the CC BY-SA 4.0
I also made a video including details & instructions.
Boosts Please! The greater the diversity of people & languages in the data, the better we can test for correctness.
PT: https://makertube.net/w/cQECfDkuLGR9eUQquUEo4K
YT: https://youtu.be/CyjumVTWtJA
Text (instructions only):
https://github.com/Helium314/HeliBoard/wiki/Tutorial:-How-to-Contribute-Gesture-Data

2026-02-05 23:17:46 UTC

- reply

are these the ones you were talking about a couple weeks ago that you had a sudden urge to make? 🙂

2026-01-03 10:49:40 UTC

- reply

Nah, not dumb, you guys just aren't following the US rat race as closely.
It's referring to Trump bombing 3 countries unprovoked, and just now sending airborne troops in to Caracas to capture Maduro...
The black woman in question is Kamala Harris. Several people cited her laugh as why they wouldn't vote for her. In reality, that was no ones' reason. It just was a plausible guise for reasons they rather wouldn't admit to, most related to apathy or bigotry.