I was speaking to a Kuwaiti princeling a few years ago about solar and he just couldn’t get his head around zero marginal cost, the efficiency of assembly of the panels, and the economics that would drive the growth. We spoke for about half an hour and he kept bringing up that powerbrokers don’t care about the environment and I had to repeatedly point out that I hadn’t mentioned the environment once.
How often are complex scripts rendered in terminal? What is the cost to scripts that are currently rendered accurately by terminal? Are there any group of tools that operate in complex scripts?
EDIT: Without saying that I think this is worthy and cool. I am just curious about the costs and benefits of such a tool.
> How often are complex scripts rendered in terminal?
If you speak the languages that use those scripts? Then all the time, I imagine. The support for double-char width cells in the terminals started to appear all the way back in the late seventies because Japan, you know, existed and kinda mattered.
All the languages I am literate in use an alphabet and I have never encountered a script in anything other alphabetic scripts in the terminal, and never anything not in English for serious work.
I would think we would probably have far fewer characters with hard to determine widths being printed in terminal (before LLMs) as most of it would be rendered in the GUI, which state of the art terminal emulators somewhat rely on anyway.
My guess is that LLMs made translation for these sorts of tools much easier (just needing someone fluent in both languages to verify rather than translate from scratch) but that's why I am asking. Is it more common now than ever before?
Beyond that the examples given were of scripts that are widely used in India which is a country with the world's largest English speaking population and one of the world's most spoken English dialects and also a huge IT sector.
I get that CJK has an existing double width carve out, that is being proposed to be kept by the objection linked in the article.
The terminal-based editors like vim/emacs exist. Ideally they should work just as well the GUI-based editors, no matter what text files they're tasked to open and edit.
As for the scripts, Unicode is slowly rounding up with the actually used scripts, but it's still not quite there yet.
And then there are emojis, of course, which we get new sets every couple of years, which makes the life of the terminal emulators simply terrible. There is simply no good way to support emoji sequences in the terminal, I believe. Consider e.g. HEART ON FIRE (READ HEART, ZWJ, FIRE) and PHOENIX (BIRD, ZWJ, FIRE) emojis: the terminal emulator may or may not be aware of the existences of those emoji sequences depending on which version of ICU it is build against, and even if it is aware of them, then the currently selected font may not have the required glyphs so it has to fall back to the fallback display (4 cells of two basic emojis vs. 2 cells of the combined glyph). And of course, the font itself may have completely bonkers dimensions for its glyphs: as I understand it, Google's Noto Emoji font has glyphs for playing cards that are 1.5 cells wide even though they're supposed to be narrow (only 1 cells wide); there is nothing much the terminal emulator can do with that.
> A common misconception is that AI is more expensive than search. Opposite is true - performing a search is 100x-1000x more expensive than doing an interaction with AI.
> A single search is about ~ 1.5 cents (probably bit less these days, but general ballpark)
> A small AI model can generate a lot of tokens for the same price
> To put this in perspective, 1.5 cents of Gemini Flash usage:
> 50,000 input tokens ≈ 37,500 words (roughly 75-150 pages of text)
or
6,000 output tokens ≈ 4,500 words (roughly 9-18 pages of text)
> That is A LOT of information for the cost of just 1 search.
> It may be counter-intuitive but that is how it is. If someone is using AI to answer common questions, learn about a concept, do a quick brainstorm or a translation - instead of searching - our cost is (much) lower.
> So if anything, having AI lowers our cost, not increases it. This is why when we added access to AI models to Pro tier we didn't increase the price.
Doesn't make sense because AI responses are not grounded, still for AI to make sense in this context, and have any relation with Kagi purpose, you need to have the search still, and then the AI process the search results.
It's worked well as a honeypot, but I don't think it's working well as a device for paralysis. The executive has seized an alarming amount of power (with the tacit approval of the party in control of the legislature), and the constitution isn't doing much of anything to stop it.
I always think that the goal of apps like these is build userbase and then get acquired, like darksky or waze: the big providers realise they have missed a trick and then it becomes the default.
"Needed tens of dollars of rental compute" isn't much of a moat to get acquired instead of copied.
I loath Waze. Its idea of shortcuts are terrible, its search routinely suggests things hundreds of mile away ahead of the match that's nearby, and sometimes I go through an area of intermittent service where it just decides to stop routing without giving a heads-up.
But Waze is so much better at accurately alerting me to police than my Valentine 1 was that I never even bothered mounting it in my latest car. Google supposedly integrates that data for years now but every time I try it comes up short. Google and Apple Maps are better in every other way, but for me at least, that one feature of Waze is a massive moat.
Nonono, despite my outward white appearance, I follow the sage wisdom of Chris Rock[0]. It's so I have enough advance notice to blast "Fuck the Fire Department" so we can share a nice laugh at the historic rivalry between LE / FD.
Yeah, being awoken by an alarm in pure darkness is grim which, longitude 15 solar noon minutes west of where our timezone is set and at our latitude is very possible in winter.
With pure standard time we would never have sunset before 5 pm but daylight savings puts half the year's 7 am before the sun has risen, and if you are an early riser as I have become, before the dawn breaks.
It also gives us four months where it's very hard to get children to sleep after 8 pm and for me it's even hard to start winding down.
I think summer time is really non-optimal for most purposes, changing the clocks sucks, and most individuals that work do so for too many hours a day. It's a local maximum in terms of how we socially manage time and people mistake optimising our society towards it to be optimising towards a global maximum.
Imagine if there was no DST and someone said "let's change every clock...", I would think it's a classic XY problem.
I have lived at high latitudes and agree; funnily enough around that time of year fog and cloud cover often meant no sun even if you were out during the day, records of 62 days with no sight of the sun. Crushing stuff.
But the situation you describe is literally not physically possible where I currently live due to proximity to the equator and being west of the line of longitude our clock runs on during standard time, but DST demands we wake up in darkness.
How the workday in the modern economy is fundamentally unjust. You shouldn't have to sign away your ability to see the sun for a job (unless at extremely high latitudes or extreme weather conditions).
Yeah, it’s interesting how the desktop metaphor evolved over time but increasing display size and the ability to have multiple workspaces surely is a huge part of what makes tiling almost work.
And tiling still largely doesn't work with small windows.
reply