Its not just JavaScript, it's cookies, it's "auto loading" resources (e.g. 1x1 pixels with per-request unique URLs), it's third-party http requests to other domains (which might art cookies too).
I think the XKCD comic about encryption-vs-wrench has never been more apt for Gemini the protocol...
Why would you only include blogs in your small web index? That must be a minute fraction of what is out there?
I can't think of a single blog that I read these days (small or not), yet there are loads of small "old school" sites out there that are still going strong.
> Why would you only include blogs in your small web index?
I am not associated with this project, so this would be a question for the project maintainer. As far as I understand, the project relies on RSS/Atom feeds to fetch new posts and display them in the search results. I believe, this is an easier problem to solve than using a full blown web crawler.
However, as far as I know, Kagi does have its own full blown crawler, so I am not entirely sure why they could not use it to present the Small Web search results. Perhaps they rely on date metadata in RSS feeds to determine whether a post was published within the last seven days? But having worked on an open source web crawler myself, many years ago, I know that this is something a web crawler can determine too if it is crawling frequently enough.
So yes, I think you have got a good point and only the project maintainer can provide a definitive answer.
Not quite. I think Gemini has deliberately gone for a "text only" philosophy, which I think is very constraining.
The early web had a lot going on and allowed for a lot of creative experimentation which really caught the eye and the imagination.
Gemini seems designed to only allow long-form text content. You can't even have a table let alone inline images which makes it very limited for even dry scientific research papers, which I think would otherwise be an excellent use-case for Gemini. But it seems that this sort of thing is a deliberate design/philosophical decision by the authors which is a shame. They could have supported full markdown, but they chose not to (ostensibly to ease client implementation but there are a squillion markdown libraries so that assertion doesn't hold water for me)
It's their protocol so they can do what they want with it, but it's why I think Gemini as a protocol is a dead-end unless all you want to do is write essays (with no images or tables or inline links or table-of-contents or MathML or SVG diagrams or anything else you can think of in markdown). Its a shame as I think the client-cert stuff for Auth is interesting.
It’s tough but one of the tenets of Gemini is that a lone programmer can write their own client in a spirited afternoon/weekend. Markdown is just a little too hard to clear the bar. Already there was much bellyaching on the mailing list about forcing dependence on SSL libraries; suggesting people rely on more libraries would have been a non-starter
Note that the Gemini protocol is just a way of moving bytes around; nothing stops you from sending Markdown if you want (and at least some clients will render it - same with inline images).
Didn't the creator of the protocol go on a rant when someone made a browser for Gemini that included a favicon?
I can't imagine the backlash if someone tried to normalize Markdown. Isn't the entire point of Gemini that it can never be extended or expanded upon?
Maybe it would be better to create an entirely different protocol/alt web around Markdown that didn't risk running afoul of Gemini's philosophical restrictions?
Yeah, instead someone makes a new and incompatible protocol whenever they want to change it.
> The SmolNet consists of content available through alternative protocols outside the web such as gemini:// gopher:// Gopher+ gophers:// finger:// spartan:// text:// SuperText nex:// scorpion:// mercury:// titan:// guppy:// scroll:// molerat:// terse:// fsp://. There is a summary of the main SmolNet protocols.
I think a "markdown-web" that uses some of the Gemini approaches for privacy and auth/identity etc would be pretty nice.
Of course, as others have said, we could just use HTML without JavaScript or cookies and we'd be a lot of the way there with 95% less effort but hey in the future we'll probably just query an AI rather than load a web page ourselves.
Given how many people on HN say they like Gemini in principle but wish it weren't so restrictive, some people would use it. All of those people might just be that cross section of HN users, however.
There are images in geminispace, and audio, and (probably) video. It's just not inline. One of constraints of the protocol is that pages cannot load content without your express say-so.
A think a more workable and politically palatable version of UBI would be some form of universal utility allowance.
E.g. the first x kWh electric you use, or the first X litres of water, or the first x GB of data you use is entirely free, for everyone (where X is some reasonable number that someone could just conceivably survive on). Then as you use more and more the prices start to gradually increase across a series of bands so that the heaviest users are subsidising those using the least.
It would promote efficiency, would be progressive, and would allow people to live without quite so much "bill fear" for essential utilities.
Plus it is not literally putting money in people's hands which is often unpopular with some demographic groups. People would still need to work but there would be some element of safety net.
> Plus it is not literally putting money in people's hands which is often unpopular with some demographic groups
I'd be really opposed to this. It'd only be ok if we nationalized the industries where we set these rules and rates. Otherwise, this ends up being a simple handout to private industries.
For example, let's say we say x liters of water. Well who's deciding how much x liters cost? If it's a private company and the government is guaranteeing it, you can bet water (which is relatively cheap where I live) will end up being the most expensive resource imaginable. And that may actually be true depending on the location, but it'd also be true in non-desert areas with plenty of water.
We've effectively had that here with the ACA, where the government has decided that it will cover the first $800 or so dollars of your health insurance. What happened? Magically, the cost of health insurance increased by $800. Private industries aren't stupid, they'll always charge the maximum price the market will bear. And when we start talking about captured industries like data provider, power provider, or your water provider... well that's where we can trust private industry the least as they literally have the public over a barrel. Utility boards are an OK solution, but the better solution is to turn these into public institutes instead of private ones.
> We've effectively had that here with the ACA, where the government has decided that it will cover the first $800 or so dollars of your health insurance. What happened? Magically, the cost of health insurance increased by $800.
I don’t think that’s an accurate description of ACA [1], it didn’t lead to a dollar to dollar increase in premiums (share a citation if otherwise), and it’s a bit misleading to say it led to an increase in premiums because plans pre-ACA were effectively inaccessible to and lacking in benefits for impoverished people or people with pre-existing conditions.
[1] Here’s a brief description of ACA from Wikipedia:
> The act largely retained the existing structure of Medicare, Medicaid, and the employer market, but individual markets were radically overhauled.[1][11] Insurers were made to accept all applicants without charging based on pre-existing conditions or demographic status (except age). To combat the resultant adverse selection, the act mandated that individuals buy insurance (or pay a monetary penalty) and that insurers cover a list of "essential health benefits". Young people were allowed to stay on their parents' insurance plans until they were 26 years old.
There will never be a cited reason for increases, but here's 2023 where basically all insurers filled for a 10% increase in premiums. [1]
Since the 2022 covid bill which significantly increased the subsidization of premiums, health insurers have found various reasons to increase their premiums by inflation beating numbers.
That's obviously a "the market will bear it" situation.
The ACA was a big bill that did a lot. I'm not talking about all of it, but rather the premium subsidization along with the covid premium increase which both expired in 2026.
Look, the premiums expiring was bad. IDK if that was clear from my earlier comment. But there's a fundamentally unaddressed issue with insurers in general where they charge not based on competition or the cost of service, but based on what consumers can bear. Profit incentives for healthcare in the US are completely misaligned with providing good general healthcare. The ACA premiums are a bandaid over an artery laceration. Better than nothing, but that thing is going to very quickly start bleeding through. You can keep slapping on band-aides, but ultimately you'll be looking at more damage if you don't just address the issue.
Utilities in America refers to the service relative to ideas of basic needs for survival in the US so they are often public infrastructure with private operators but in the case of some things like the internet, it’s purely privatized.
A non-tradable utility allowance would incent people to waste these basic goods, and a tradable utility allowance, while obviously fixing this, would be no different than a UBI that was indexed to the price of basic utilities.
That seems worse because it doesn't encourage conservation enough. You want people to be able to keep the gains if they conserve energy, to set the right incentives.
The “demographic groups” that find it unpopular dislike anything that can be remotely construed as giving to the poor. There is literally no purpose trying to cater to them while trying to set up a social program.
> prioritization and decision frameworks start to matter more.
This is the thing though, prioritization doesn't matter in the same way it used to.
We only needed to prioritize before because engineering was relatively slow and precious resource, so we had to pick and chose what to work on first because it took time.
But now we effectively have a limitless supply of SWEs, so why not do everything on the backlog?
I think the question now is more about sequencing than prioritization. What do we need to do first, before we can do these other things?
But yes generally requirements are still very important. Which features do we need etc.
I disagree. There is some argument to be had that they're already generally intelligent. They're already certainly better than me in basically anything I can ask them to do.
So that leads to the question of what qualifies as intelligent? And do we need sentience for intelligence? What about self-agency/-actuation? Is that needed for "generally intelligent"?
I don't know.
But I feel like we're not there yet, even for non-sentient intelligence. I personally think we need an "unlimited" context (as good as human memory context windows anyway, which some argue we've already surpassed) and genuine self-learning before we get close. I don't think we need it to be an infallible genius (i.e
ASI) to qualify as generally intelligent ... or to put it another way "about as smart and reliable as the average human adult" which frankly is quite a low bar!
One thing for sure though, I think this will creep up on us and one day it will suddenly become apparent that it's already there and we just didn't appreciate/notice/comprehend. There won't be a big fireworks display the moment it happens, more of a creeping realisation I think.
How much of our brain's innate wiring is "pre-training"? We're born pre-wired to breath, to swallow, to blink, sleep, cry etc. Fine tuned over many many many epochs and baked into our model weights/DNA.
Is a newborn baby without learnt-knowledge not an intelligent being to you?
Or is an empty vessel such as a newborn baby intelligent merely because it has the ability to learn?
It gets pretty philosophical pretty quick. This is why I don't think there'll be a "moment" when AGI happens - there is so many ways to interpret what constitutes intelligence.
But yeah I agree that until models can learn in real time then I think we're probably not there yet. As I said - 5 years give or take I reckon.
> Is a newborn baby without learnt-knowledge not an intelligent being to you?
A newborn baby without learnt knowledge is a phenomenal comparison to an LLM: reactionary, incapable of communicating a novel thought, extremely inconsistent reactions to similar stimuli, costs a lot of money, and the best part is they almost never turn out to be an income-bearing investment.
Despite all this, people vehemently defend their ugly, obnoxious, screaming, drooling babies with the fierceness of a lion, because they’re so blinded by emotion they’re incapable of logical thought.
I have three kids, I’ve earned the right to say this.
+1 In any major city it's probably 90% chance they're either a crook trying to scam you out of something or mentally not quite right. The remaining 10% will be tourists or people from outside of the major city.
"Someday - and that day may never come" - they will be asking for something. Now you are on talking terms and it will be harder to refuse the ask, compared to the request of a complete stranger.
This is sad but not inconsistent with my experience. Though I think 10% is actually the people who genuinely want to have a nice conversation. and I think that worth putting up witrh the rest 90% for.
I understand what both of you are saying, I lived in areas where if someone is talking to you on the street theres a high chance theyre asking you for something, so you learn to just kinda block all of it out. Now that I moved to a smaller town, I find myself talking to strangers much more frequently.
Its not just JavaScript, it's cookies, it's "auto loading" resources (e.g. 1x1 pixels with per-request unique URLs), it's third-party http requests to other domains (which might art cookies too).
I think the XKCD comic about encryption-vs-wrench has never been more apt for Gemini the protocol...
reply