savolai

I would suggest having the place button next to/in the item being dragged like Pick. When learning to use this, having it at a distance creates unnecessary cognitive burden.

Also, I would probably make the entire item clickable for pick, if there didn’t have to be a click target on the item for other functionality.

https://marshallbrain.com/manna1

Social work demands judgment, trust, and moral accountability. Teachers shape citizens, not outputs. Health care relies on responsibility under uncertainty, not pattern matching. States run on legitimacy and human obligation, not efficiency alone.

Even if AI boosts productivity across these sectors, someone still funds universal care, education, and social insurance. Remove taxes and the welfare state collapses long before any ”AI government” appears.

Claiming AI replaces government skips the core question: who guarantees care when systems fail, people fall through cracks, or outcomes turn unjust? Algorithms can advise. They can’t answer politically or ethically.

Tax the concentration of wealth and power they amplify. Without that, automation doesn’t liberate society. It hollows it out.

Meanwhile, one of the linked pages mentioned bluegriffon, so I got curious if such editors can handle template languages such as django's.

While I don't know that yet (edit:no templates at all, what a shame), I also found this tutorial, and it was inspiring that such a pedagogical approach to online authoring still exists.

https://www.thesitewizard.com/bluegriffon/bluegriffon-2-tuto...

Edit: no seriously, why don't these editors support at least some established template language? I think dreamweaver had a concept of templates, which made using these editors make at least some sense.

Edit: oh wow dreamweaver still exists. Any of you have experience? Still good?

So various projects have come up ever since to try to patch the UX.

The single most valuable tool is user testing. However it really takes quite a few rounds of actually creating a design and seeing how wrong you saw the other person’s capabilities, to grok how powerful user testing is in revealing your own biases.

And it’s not hard at all at core. The most important lesson really is a bit of humility. Actually shutting up and observing what real users do when not intervened.

Shameless plug, my intro to user testing: https://savolai.net/ux/the-why-and-the-how-usability-testing...

The idea that “we just need herd immunity” assumes we’re all equally capable of recognizing manipulation or addiction, but as ericmcer noted, the evidence all around us suggests most of us are not quite there. In many ways, that belief in our individual mastery is part of what Western culture keeps overestimating, as if understanding the trick is enough to undo the conditioning.

There also seems to be a deeper resistance in our western cultures to actually pausing, turning inward, and staying present with what’s happening inside us. We intellectualize awareness instead of living it. Real freedom will begin not with more information or clever systems, but with learning to meet our own impulses directly. To listen, to stay with discomfort, to see what drives us without immediately trying to fix it. Until we're willing to practice that kind of contact with ourselves, we’ll keep playing defense against the surface symptoms.

Obligatory Manna link

I wondered could this be the architecture i’m looking for since I want something that is offline first but works online. Hyperclay removes a lot of the architectural heft of traditional web apps. But of course now the data is in the dom. So xslt could in theory be used to bring it back into an actually-structural format.

I.e. treat the DOM as a view and use XSLT as the lossless bridge back to structured, declarative and deterministic config. Round-trips are doable. One stylesheet renders config → HTML editor. Another extracts edited DOM → config.

Of course, if downstream systems expected JSON, transforming DOM→JSON via JS might be simpler. My previous backend is sqlite though.

I would like to have more of the content available offline automatically though, i.e all text and image content, and big files downloaded on request. Closer to local first than this.

To me it seems that developer/tinkerer types strongly live in an echo chamber. Of course we are both speculating here, but to me it seems that's exacly where apple derives its market value. By emphasizing the needs of designer types and "ordinary people" in contrast to techies.

The former don't necessarily derive lots of value from, say, having real file systems, which tinkerers often want.

I would claim ordinary people buy android mainly due to price. (Ofc there are also premium android phones where change resistance and pure marketing on both sides of the fence may affect decisions more)

I would propose that those strongly committed to learning tech, rarely see the amount of work they have put into learning and tweaking the system. They do not perceive it as work but as a) having learnt "general knowledge" and b) something they want to do as a matter of fact.

Product/framework thinkers is another way to think of this: https://savolai.net/ux/product-and-framework-thinkers-when-d...

Those people often, unsurprisingly, are also developers.

So those who want to focus on multitude of other things in life and don't want to invest so much in dev tech, don't have their voices heard in dev communities. I have always felt very marginalized here. (Please don't read this wrong, tinkerers absolutely also enjoy a "multitude of things". The emphasis is just perhaps more specialized in a specifically weighted way.)

The "hidden controls" link you provide is interesting, as it could easily be used as an argument against the original linux terminal copy paste issues too. Terminal use in and of itself is of course an example of everything being hidden by default.

It seems in touch interfaces no one can completely avoid this, or at least the industry has strongly moved away from complete visibility/affordances. Which is kinda fascinating, in such visual medium. I would love a physical keyboard but can see why that didn't pan out.

I’m a developer (fullstack, conceptual modeling, db, architecture, c++/qt, php, python, cs degree) who is trained in UX too and using windows and linux is painful because I never get to enjoy the customization part due the poor nature of default UX ie results of bikeshedding everywhere.

It is just too much work to get to basic ok defaults, to have any energy left to think about what I might want to customize.

The system forcing users to customize is just as much use of power as not being able to.

Of course the ideal is progressive disclosure i.e easy things easy (good defaults) and hard things possible (a configuration dialog).

(I’ve written a brief intro to progressive disclosure here: https://savolai.net/ux/ui-design-balancing-user-needs-with-p... )

But the line has to be drawn somewhere, as apple has. Being able to bikeshed and customize anything can easily become a multiplier of complexity and maintenance cost. It’s not any less opinionated than wanting to keep things reasonably simple.

For some it’s a reasonable tradeoff, for others not. For me the value of apple consistency and aesthetics far outweigh the costs. Sure, there is a learning curve and change resistance I had to go through coming from win/linux, but I wouldn’t say macos has any significant barriers to what I can do. Quite the contrary with M class cpus.

Iphones are another story, but eventually the tradeoffs outweigh android ux illogical nature and inconsistency there too.

I am already cognitively burdened. I do not need developers telling me at what point the cumulative inconsistencies become a biggie. Copy-paste is a common action and each inconsistency I need to learn is away from my core tasks and ability to focus on those is already scarce as f.

Devs do not get to decide how central a terminal is to my workflow, and whether that terminal app deserves to have the right to tell me that it’s now a special butterfly I need to accommodate my cognition for.

But I guess Linux desktop has chosen its path of being only for tinkerers who are prepared to adopt an entire culture of quirks instead of users focusing on what’s important for them in their own lives.

I’m disappointed this still does not fix the core issue of this being broken for everyone by default.

Economic Times seems to have drawn upside-down conclusions. The whole article comes off as alarmist at first glance.

As far as I understand, the actual “thinking process” of LLMs—statistical language generation—has never really been transparent.

Chain of Thought, which the original paper discusses, adds a new layer on top of language generation. It tries to make the underlying process more visible by “slowing it down,” breaking it into smaller steps. It’s been around for a long time.

To my understanding, CoT is also, in a way, an appearance—because the actual reasoning of the AI remains hidden behind statistical mechanisms. Apparently, though, it reduces errors.

At least based on the summary, the original paper doesn’t seem to be “warning” about anything.

I don’t get why Chain of Thought—the method is now considered some novel idea worth special news coverage, or why transparency in reasoning would be especially under threat right now.

This way user could directly review if it is suggesting something they want to go on with.